Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / python / google-api-python-client / f80be181200f4aae910badec3f4bddea08ae5044 / . / docs / dyn / bigquery_v2.models.html
blob: 2f813fe8b4d22ed6c44bedcbaf7fd1745c1663ac [file] [log] [blame]
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1<html><body>
2<style>
3
4body, h1, h2, h3, div, span, p, pre, a {
5 margin: 0;
6 padding: 0;
7 border: 0;
8 font-weight: inherit;
9 font-style: inherit;
10 font-size: 100%;
11 font-family: inherit;
12 vertical-align: baseline;
13}
14
15body {
16 font-size: 13px;
17 padding: 1em;
18}
19
20h1 {
21 font-size: 26px;
22 margin-bottom: 1em;
23}
24
25h2 {
26 font-size: 24px;
27 margin-bottom: 1em;
28}
29
30h3 {
31 font-size: 20px;
32 margin-bottom: 1em;
33 margin-top: 1em;
34}
35
36pre, code {
37 line-height: 1.5;
38 font-family: Monaco, 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Lucida Console', monospace;
39}
40
41pre {
42 margin-top: 0.5em;
43}
44
45h1, h2, h3, p {
46 font-family: Arial, sans serif;
47}
48
49h1, h2, h3 {
50 border-bottom: solid #CCC 1px;
51}
52
53.toc_element {
54 margin-top: 0.5em;
55}
56
57.firstline {
58 margin-left: 2 em;
59}
60
61.method {
62 margin-top: 1em;
63 border: solid 1px #CCC;
64 padding: 1em;
65 background: #EEE;
66}
67
68.details {
69 font-weight: bold;
70 font-size: 14px;
71}
72
73</style>
74
75<h1><a href="bigquery_v2.html">BigQuery API</a> . <a href="bigquery_v2.models.html">models</a></h1>
76<h2>Instance Methods</h2>
77<p class="toc_element">
78 <code><a href="#delete">delete(projectId, datasetId, modelId)</a></code></p>
79<p class="firstline">Deletes the model specified by modelId from the dataset.</p>
80<p class="toc_element">
81 <code><a href="#get">get(projectId, datasetId, modelId)</a></code></p>
82<p class="firstline">Gets the specified model resource by model ID.</p>
83<p class="toc_element">
84 <code><a href="#list">list(projectId, datasetId, pageToken=None, maxResults=None)</a></code></p>
85<p class="firstline">Lists all models in the specified dataset. Requires the READER dataset</p>
86<p class="toc_element">
87 <code><a href="#list_next">list_next(previous_request, previous_response)</a></code></p>
88<p class="firstline">Retrieves the next page of results.</p>
89<p class="toc_element">
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]90 <code><a href="#patch">patch(projectId, datasetId, modelId, body=None)</a></code></p>
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]91<p class="firstline">Patch specific fields in the specified model.</p>
92<h3>Method Details</h3>
93<div class="method">
94 <code class="details" id="delete">delete(projectId, datasetId, modelId)</code>
95 <pre>Deletes the model specified by modelId from the dataset.
96
97Args:
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]98 projectId: string, Required. Project ID of the model to delete. (required)
99 datasetId: string, Required. Dataset ID of the model to delete. (required)
100 modelId: string, Required. Model ID of the model to delete. (required)
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]101</pre>
102</div>
103
104<div class="method">
105 <code class="details" id="get">get(projectId, datasetId, modelId)</code>
106 <pre>Gets the specified model resource by model ID.
107
108Args:
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]109 projectId: string, Required. Project ID of the requested model. (required)
110 datasetId: string, Required. Dataset ID of the requested model. (required)
111 modelId: string, Required. Model ID of the requested model. (required)
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]112
113Returns:
114 An object of the form:
115
116 {
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]117 "labels": { # The labels associated with this model. You can use these to organize
118 # and group your models. Label keys and values can be no longer
119 # than 63 characters, can only contain lowercase letters, numeric
120 # characters, underscores and dashes. International characters are allowed.
121 # Label values are optional. Label keys must start with a letter and each
122 # label in the list must have a different key.
123 "a_key": "A String",
124 },
125 "description": "A String", # Optional. A user-friendly description of this model.
126 "trainingRuns": [ # Output only. Information for all training runs in increasing order of start_time.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]127 { # Information about a single training query run for the model.
128 "evaluationMetrics": { # Evaluation metrics of a model. These are either computed on all training # The evaluation metrics over training/eval data that were computed at the
129 # end of training.
130 # data or just the eval data based on whether eval data was used during
131 # training. These are not present for imported models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]132 "clusteringMetrics": { # Evaluation metrics for clustering models. # Populated for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]133 "meanSquaredDistance": 3.14, # Mean of squared distances between each sample to its cluster centroid.
134 "daviesBouldinIndex": 3.14, # Davies-Bouldin index.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]135 "clusters": [ # [Beta] Information for all clusters.
136 { # Message containing the information about one cluster.
137 "count": "A String", # Count of training data rows that were assigned to this cluster.
138 "featureValues": [ # Values of highly variant features for this cluster.
139 { # Representative value of a single feature within the cluster.
140 "featureColumn": "A String", # The feature column name.
141 "numericalValue": 3.14, # The numerical feature value. This is the centroid value for this
142 # feature.
143 "categoricalValue": { # Representative value of a categorical feature. # The categorical feature value.
144 "categoryCounts": [ # Counts of all categories for the categorical feature. If there are
145 # more than ten categories, we return top ten (by count) and return
146 # one more CategoryCount with category "_OTHER_" and count as
147 # aggregate counts of remaining categories.
148 { # Represents the count of a single category within the cluster.
149 "category": "A String", # The name of category.
150 "count": "A String", # The count of training samples matching the category within the
151 # cluster.
152 },
153 ],
154 },
155 },
156 ],
157 "centroidId": "A String", # Centroid id.
158 },
159 ],
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]160 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]161 "regressionMetrics": { # Evaluation metrics for regression and explicit feedback type matrix # Populated for regression models and explicit feedback type matrix
162 # factorization models.
163 # factorization models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]164 "meanSquaredLogError": 3.14, # Mean squared log error.
165 "meanAbsoluteError": 3.14, # Mean absolute error.
166 "meanSquaredError": 3.14, # Mean squared error.
167 "medianAbsoluteError": 3.14, # Median absolute error.
168 "rSquared": 3.14, # R^2 score.
169 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]170 "rankingMetrics": { # Evaluation metrics used by weighted-ALS models specified by # [Alpha] Populated for implicit feedback type matrix factorization
171 # models.
172 # feedback_type=implicit.
173 "meanSquaredError": 3.14, # Similar to the mean squared error computed in regression and explicit
174 # recommendation models except instead of computing the rating directly,
175 # the output from evaluate is computed against a preference which is 1 or 0
176 # depending on if the rating exists or not.
177 "meanAveragePrecision": 3.14, # Calculates a precision per user for all the items by ranking them and
178 # then averages all the precisions across all the users.
179 "averageRank": 3.14, # Determines the goodness of a ranking by computing the percentile rank
180 # from the predicted confidence and dividing it by the original rank.
181 "normalizedDiscountedCumulativeGain": 3.14, # A metric to determine the goodness of a ranking calculated from the
182 # predicted confidence by comparing it to an ideal rank measured by the
183 # original ratings.
184 },
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]185 "binaryClassificationMetrics": { # Evaluation metrics for binary classification/classifier models. # Populated for binary classification/classifier models.
186 "negativeLabel": "A String", # Label representing the negative class.
187 "aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.
188 # models, the metrics are either macro-averaged or micro-averaged. When
189 # macro-averaged, the metrics are calculated for each label and then an
190 # unweighted average is taken of those values. When micro-averaged, the
191 # metric is calculated globally by counting the total number of correctly
192 # predicted rows.
193 "recall": 3.14, # Recall is the fraction of actual positive labels that were given a
194 # positive prediction. For multiclass this is a macro-averaged metric.
195 "precision": 3.14, # Precision is the fraction of actual positive predictions that had
196 # positive actual labels. For multiclass this is a macro-averaged
197 # metric treating each class as a binary classifier.
198 "logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.
199 "threshold": 3.14, # Threshold at which the metrics are computed. For binary
200 # classification models this is the positive class threshold.
201 # For multi-class classfication models this is the confidence
202 # threshold.
203 "accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For
204 # multiclass this is a micro-averaged metric.
205 "f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass
206 # this is a macro-averaged metric.
207 "rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged
208 # metric.
209 },
210 "positiveLabel": "A String", # Label representing the positive class.
211 "binaryConfusionMatrixList": [ # Binary confusion matrix at multiple thresholds.
212 { # Confusion matrix for binary classification models.
213 "truePositives": "A String", # Number of true samples predicted as true.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]214 "recall": 3.14, # The fraction of actual positive labels that were given a positive
215 # prediction.
216 "precision": 3.14, # The fraction of actual positive predictions that had positive actual
217 # labels.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]218 "falseNegatives": "A String", # Number of false samples predicted as false.
219 "trueNegatives": "A String", # Number of true samples predicted as false.
220 "falsePositives": "A String", # Number of false samples predicted as true.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]221 "f1Score": 3.14, # The equally weighted average of recall and precision.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]222 "positiveClassThreshold": 3.14, # Threshold value used when computing each of the following metric.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]223 "accuracy": 3.14, # The fraction of predictions given the correct label.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]224 },
225 ],
226 },
227 "multiClassClassificationMetrics": { # Evaluation metrics for multi-class classification/classifier models. # Populated for multi-class classification/classifier models.
228 "aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.
229 # models, the metrics are either macro-averaged or micro-averaged. When
230 # macro-averaged, the metrics are calculated for each label and then an
231 # unweighted average is taken of those values. When micro-averaged, the
232 # metric is calculated globally by counting the total number of correctly
233 # predicted rows.
234 "recall": 3.14, # Recall is the fraction of actual positive labels that were given a
235 # positive prediction. For multiclass this is a macro-averaged metric.
236 "precision": 3.14, # Precision is the fraction of actual positive predictions that had
237 # positive actual labels. For multiclass this is a macro-averaged
238 # metric treating each class as a binary classifier.
239 "logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.
240 "threshold": 3.14, # Threshold at which the metrics are computed. For binary
241 # classification models this is the positive class threshold.
242 # For multi-class classfication models this is the confidence
243 # threshold.
244 "accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For
245 # multiclass this is a micro-averaged metric.
246 "f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass
247 # this is a macro-averaged metric.
248 "rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged
249 # metric.
250 },
251 "confusionMatrixList": [ # Confusion matrix at different thresholds.
252 { # Confusion matrix for multi-class classification models.
253 "confidenceThreshold": 3.14, # Confidence threshold used when computing the entries of the
254 # confusion matrix.
255 "rows": [ # One row per actual label.
256 { # A single row in the confusion matrix.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]257 "actualLabel": "A String", # The original label of this row.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]258 "entries": [ # Info describing predicted label distribution.
259 { # A single entry in the confusion matrix.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]260 "predictedLabel": "A String", # The predicted label. For confidence_threshold &gt; 0, we will
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]261 # also add an entry indicating the number of items under the
262 # confidence threshold.
263 "itemCount": "A String", # Number of items being predicted as this label.
264 },
265 ],
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]266 },
267 ],
268 },
269 ],
270 },
271 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]272 "dataSplitResult": { # Data split result. This contains references to the training and evaluation # Data split result of the training run. Only set when the input data is
273 # actually split.
274 # data tables that were used to train the model.
275 "trainingTable": { # Table reference of the training data after split.
276 "projectId": "A String", # [Required] The ID of the project containing this table.
277 "tableId": "A String", # [Required] The ID of the table. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
278 "datasetId": "A String", # [Required] The ID of the dataset containing this table.
279 },
280 "evaluationTable": { # Table reference of the evaluation data after split.
281 "projectId": "A String", # [Required] The ID of the project containing this table.
282 "tableId": "A String", # [Required] The ID of the table. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
283 "datasetId": "A String", # [Required] The ID of the dataset containing this table.
284 },
285 },
286 "results": [ # Output of each iteration run, results.size() &lt;= max_iterations.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]287 { # Information about a single iteration of the training run.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]288 "arimaResult": { # (Auto-)arima fitting result. Wrap everything in ArimaResult for easier
289 # refactoring if we want to use model-specific iteration results.
290 "arimaModelInfo": [ # This message is repeated because there are multiple arima models
291 # fitted in auto-arima. For non-auto-arima model, its size is one.
292 { # Arima model information.
293 "seasonalPeriods": [ # Seasonal periods. Repeated because multiple periods are supported
294 # for one time series.
295 "A String",
296 ],
297 "hasDrift": True or False, # Whether Arima model fitted with drift or not. It is always false
298 # when d is not 1.
299 "arimaCoefficients": { # Arima coefficients. # Arima coefficients.
300 "movingAverageCoefficients": [ # Moving-average coefficients, an array of double.
301 3.14,
302 ],
303 "autoRegressiveCoefficients": [ # Auto-regressive coefficients, an array of double.
304 3.14,
305 ],
306 "interceptCoefficient": 3.14, # Intercept coefficient, just a double not an array.
307 },
308 "nonSeasonalOrder": { # Arima order, can be used for both non-seasonal and seasonal parts. # Non-seasonal order.
309 "q": "A String", # Order of the moving-average part.
310 "p": "A String", # Order of the autoregressive part.
311 "d": "A String", # Order of the differencing part.
312 },
313 "arimaFittingMetrics": { # ARIMA model fitting metrics. # Arima fitting metrics.
314 "variance": 3.14, # Variance.
315 "logLikelihood": 3.14, # Log-likelihood.
316 "aic": 3.14, # AIC.
317 },
318 "timeSeriesId": "A String", # The id to indicate different time series.
319 },
320 ],
321 "seasonalPeriods": [ # Seasonal periods. Repeated because multiple periods are supported for
322 # one time series.
323 "A String",
324 ],
325 },
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]326 "index": 42, # Index of the iteration, 0 based.
327 "evalLoss": 3.14, # Loss computed on the eval data at the end of iteration.
328 "durationMs": "A String", # Time taken to run the iteration in milliseconds.
329 "learnRate": 3.14, # Learn rate used for this iteration.
330 "trainingLoss": 3.14, # Loss computed on the training data at the end of iteration.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]331 "clusterInfos": [ # Information about top clusters for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]332 { # Information about a single cluster for clustering model.
333 "centroidId": "A String", # Centroid id.
334 "clusterSize": "A String", # Cluster size, the total number of points assigned to the cluster.
335 "clusterRadius": 3.14, # Cluster radius, the average distance from centroid
336 # to each point assigned to the cluster.
337 },
338 ],
339 },
340 ],
341 "startTime": "A String", # The start time of this training run.
342 "trainingOptions": { # Options that were used for this training run, includes
343 # user specified and default options that were used.
344 "optimizationStrategy": "A String", # Optimization strategy for training linear regression models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]345 "itemColumn": "A String", # Item column specified for matrix factorization models.
346 "feedbackType": "A String", # Feedback type that specifies which algorithm to run for matrix
347 # factorization.
348 "numFactors": "A String", # Num factors specified for matrix factorization models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]349 "inputLabelColumns": [ # Name of input label columns in training data.
350 "A String",
351 ],
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]352 "batchSize": "A String", # Batch size for dnn models.
353 "distanceType": "A String", # Distance type for clustering models.
354 "kmeansInitializationColumn": "A String", # The column used to provide the initial centroids for kmeans algorithm
355 # when kmeans_initialization_method is CUSTOM.
356 "l2Regularization": 3.14, # L2 regularization coefficient.
357 "dropout": 3.14, # Dropout probability for dnn models.
358 "minRelativeProgress": 3.14, # When early_stop is true, stops training when accuracy improvement is
359 # less than 'min_relative_progress'. Used only for iterative training
360 # algorithms.
361 "l1Regularization": 3.14, # L1 regularization coefficient.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]362 "maxIterations": "A String", # The maximum number of iterations in training. Used only for iterative
363 # training algorithms.
364 "earlyStop": True or False, # Whether to stop early when the loss doesn't improve significantly
365 # any more (compared to min_relative_progress). Used only for iterative
366 # training algorithms.
367 "initialLearnRate": 3.14, # Specifies the initial learning rate for the line search learn rate
368 # strategy.
369 "dataSplitColumn": "A String", # The column to split data with. This column won't be used as a
370 # feature.
371 # 1. When data_split_method is CUSTOM, the corresponding column should
372 # be boolean. The rows with true value tag are eval data, and the false
373 # are training data.
374 # 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION
375 # rows (from smallest to largest) in the corresponding column are used
376 # as training data, and the rest are eval data. It respects the order
377 # in Orderable data types:
378 # https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data-type-properties
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]379 "numClusters": "A String", # Number of clusters for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]380 "warmStart": True or False, # Whether to train a model from the last checkpoint.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]381 "hiddenUnits": [ # Hidden units for dnn models.
382 "A String",
383 ],
384 "maxTreeDepth": "A String", # Maximum depth of a tree for boosted tree models.
385 "userColumn": "A String", # User column specified for matrix factorization models.
386 "kmeansInitializationMethod": "A String", # The method used to initialize the centroids for kmeans algorithm.
387 "learnRateStrategy": "A String", # The strategy to determine learn rate for the current iteration.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]388 "dataSplitEvalFraction": 3.14, # The fraction of evaluation data over the whole input data. The rest
389 # of data will be used as training data. The format should be double.
390 # Accurate to two decimal places.
391 # Default value is 0.2.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]392 "dataSplitMethod": "A String", # The data split type for training and evaluation, e.g. RANDOM.
393 "subsample": 3.14, # Subsample fraction of the training data to grow tree to prevent
394 # overfitting for boosted tree models.
395 "labelClassWeights": { # Weights associated with each label class, for rebalancing the
396 # training data. Only applicable for classification models.
397 "a_key": 3.14,
398 },
399 "learnRate": 3.14, # Learning rate in training. Used only for iterative training algorithms.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]400 "modelUri": "A String", # [Beta] Google Cloud Storage URI from which the model was imported. Only
401 # applicable for imported models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]402 "walsAlpha": 3.14, # Hyperparameter for matrix factoration when implicit feedback type is
403 # specified.
404 "minSplitLoss": 3.14, # Minimum split loss for boosted tree models.
405 "lossType": "A String", # Type of loss function used during training run.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]406 },
407 },
408 ],
409 "featureColumns": [ # Output only. Input feature columns that were used to train this model.
410 { # A field or a column.
411 "type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly
412 # specified (e.g., CREATE FUNCTION statement can omit the return type;
413 # in this case the output parameter does not have this "type" field).
414 # Examples:
415 # INT64: {type_kind="INT64"}
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]416 # ARRAY&lt;STRING&gt;: {type_kind="ARRAY", array_element_type="STRING"}
417 # STRUCT&lt;x STRING, y ARRAY&lt;DATE&gt;&gt;:
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]418 # {type_kind="STRUCT",
419 # struct_type={fields=[
420 # {name="x", type={type_kind="STRING"}},
421 # {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}
422 # ]}}
423 "structType": { # The fields of this struct, in order, if type_kind = "STRUCT".
424 "fields": [
425 # Object with schema name: StandardSqlField
426 ],
427 },
428 "arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".
429 "typeKind": "A String", # Required. The top level type of this field.
430 # Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").
431 },
432 "name": "A String", # Optional. The name of this field. Can be absent for struct fields.
433 },
434 ],
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]435 "labelColumns": [ # Output only. Label columns that were used to train this model.
436 # The output of the model will have a "predicted_" prefix to these columns.
437 { # A field or a column.
438 "type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly
439 # specified (e.g., CREATE FUNCTION statement can omit the return type;
440 # in this case the output parameter does not have this "type" field).
441 # Examples:
442 # INT64: {type_kind="INT64"}
443 # ARRAY&lt;STRING&gt;: {type_kind="ARRAY", array_element_type="STRING"}
444 # STRUCT&lt;x STRING, y ARRAY&lt;DATE&gt;&gt;:
445 # {type_kind="STRUCT",
446 # struct_type={fields=[
447 # {name="x", type={type_kind="STRING"}},
448 # {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}
449 # ]}}
450 "structType": { # The fields of this struct, in order, if type_kind = "STRUCT".
451 "fields": [
452 # Object with schema name: StandardSqlField
453 ],
454 },
455 "arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".
456 "typeKind": "A String", # Required. The top level type of this field.
457 # Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").
458 },
459 "name": "A String", # Optional. The name of this field. Can be absent for struct fields.
460 },
461 ],
462 "creationTime": "A String", # Output only. The time when this model was created, in millisecs since the epoch.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]463 "modelType": "A String", # Output only. Type of the model resource.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]464 "encryptionConfiguration": { # Custom encryption configuration (e.g., Cloud KMS keys). This shows the
465 # encryption configuration of the model data while stored in BigQuery
466 # storage. This field can be used with PatchModel to update encryption key
467 # for an already encrypted model.
468 "kmsKeyName": "A String", # [Optional] Describes the Cloud KMS encryption key that will be used to protect destination BigQuery table. The BigQuery Service Account associated with your project requires access to this encryption key.
469 },
470 "modelReference": { # Required. Unique identifier for this model.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]471 "projectId": "A String", # [Required] The ID of the project containing this model.
472 "datasetId": "A String", # [Required] The ID of the dataset containing this model.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]473 "modelId": "A String", # [Required] The ID of the model. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]474 },
475 "etag": "A String", # Output only. A hash of this resource.
476 "location": "A String", # Output only. The geographic location where the model resides. This value
477 # is inherited from the dataset.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]478 "friendlyName": "A String", # Optional. A descriptive name for this model.
479 "expirationTime": "A String", # Optional. The time when this model expires, in milliseconds since the epoch.
480 # If not present, the model will persist indefinitely. Expired models
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]481 # will be deleted and their storage reclaimed. The defaultTableExpirationMs
482 # property of the encapsulating dataset can be used to set a default
483 # expirationTime on newly created models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]484 "lastModifiedTime": "A String", # Output only. The time when this model was last modified, in millisecs since the epoch.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]485 }</pre>
486</div>
487
488<div class="method">
489 <code class="details" id="list">list(projectId, datasetId, pageToken=None, maxResults=None)</code>
490 <pre>Lists all models in the specified dataset. Requires the READER dataset
491role.
492
493Args:
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]494 projectId: string, Required. Project ID of the models to list. (required)
495 datasetId: string, Required. Dataset ID of the models to list. (required)
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]496 pageToken: string, Page token, returned by a previous call to request the next page of
497results
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]498 maxResults: integer, The maximum number of results to return in a single response page.
499Leverage the page tokens to iterate through the entire collection.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]500
501Returns:
502 An object of the form:
503
504 {
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]505 "nextPageToken": "A String", # A token to request the next page of results.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]506 "models": [ # Models in the requested dataset. Only the following fields are populated:
507 # model_reference, model_type, creation_time, last_modified_time and
508 # labels.
509 {
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]510 "labels": { # The labels associated with this model. You can use these to organize
511 # and group your models. Label keys and values can be no longer
512 # than 63 characters, can only contain lowercase letters, numeric
513 # characters, underscores and dashes. International characters are allowed.
514 # Label values are optional. Label keys must start with a letter and each
515 # label in the list must have a different key.
516 "a_key": "A String",
517 },
518 "description": "A String", # Optional. A user-friendly description of this model.
519 "trainingRuns": [ # Output only. Information for all training runs in increasing order of start_time.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]520 { # Information about a single training query run for the model.
521 "evaluationMetrics": { # Evaluation metrics of a model. These are either computed on all training # The evaluation metrics over training/eval data that were computed at the
522 # end of training.
523 # data or just the eval data based on whether eval data was used during
524 # training. These are not present for imported models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]525 "clusteringMetrics": { # Evaluation metrics for clustering models. # Populated for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]526 "meanSquaredDistance": 3.14, # Mean of squared distances between each sample to its cluster centroid.
527 "daviesBouldinIndex": 3.14, # Davies-Bouldin index.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]528 "clusters": [ # [Beta] Information for all clusters.
529 { # Message containing the information about one cluster.
530 "count": "A String", # Count of training data rows that were assigned to this cluster.
531 "featureValues": [ # Values of highly variant features for this cluster.
532 { # Representative value of a single feature within the cluster.
533 "featureColumn": "A String", # The feature column name.
534 "numericalValue": 3.14, # The numerical feature value. This is the centroid value for this
535 # feature.
536 "categoricalValue": { # Representative value of a categorical feature. # The categorical feature value.
537 "categoryCounts": [ # Counts of all categories for the categorical feature. If there are
538 # more than ten categories, we return top ten (by count) and return
539 # one more CategoryCount with category "_OTHER_" and count as
540 # aggregate counts of remaining categories.
541 { # Represents the count of a single category within the cluster.
542 "category": "A String", # The name of category.
543 "count": "A String", # The count of training samples matching the category within the
544 # cluster.
545 },
546 ],
547 },
548 },
549 ],
550 "centroidId": "A String", # Centroid id.
551 },
552 ],
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]553 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]554 "regressionMetrics": { # Evaluation metrics for regression and explicit feedback type matrix # Populated for regression models and explicit feedback type matrix
555 # factorization models.
556 # factorization models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]557 "meanSquaredLogError": 3.14, # Mean squared log error.
558 "meanAbsoluteError": 3.14, # Mean absolute error.
559 "meanSquaredError": 3.14, # Mean squared error.
560 "medianAbsoluteError": 3.14, # Median absolute error.
561 "rSquared": 3.14, # R^2 score.
562 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]563 "rankingMetrics": { # Evaluation metrics used by weighted-ALS models specified by # [Alpha] Populated for implicit feedback type matrix factorization
564 # models.
565 # feedback_type=implicit.
566 "meanSquaredError": 3.14, # Similar to the mean squared error computed in regression and explicit
567 # recommendation models except instead of computing the rating directly,
568 # the output from evaluate is computed against a preference which is 1 or 0
569 # depending on if the rating exists or not.
570 "meanAveragePrecision": 3.14, # Calculates a precision per user for all the items by ranking them and
571 # then averages all the precisions across all the users.
572 "averageRank": 3.14, # Determines the goodness of a ranking by computing the percentile rank
573 # from the predicted confidence and dividing it by the original rank.
574 "normalizedDiscountedCumulativeGain": 3.14, # A metric to determine the goodness of a ranking calculated from the
575 # predicted confidence by comparing it to an ideal rank measured by the
576 # original ratings.
577 },
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]578 "binaryClassificationMetrics": { # Evaluation metrics for binary classification/classifier models. # Populated for binary classification/classifier models.
579 "negativeLabel": "A String", # Label representing the negative class.
580 "aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.
581 # models, the metrics are either macro-averaged or micro-averaged. When
582 # macro-averaged, the metrics are calculated for each label and then an
583 # unweighted average is taken of those values. When micro-averaged, the
584 # metric is calculated globally by counting the total number of correctly
585 # predicted rows.
586 "recall": 3.14, # Recall is the fraction of actual positive labels that were given a
587 # positive prediction. For multiclass this is a macro-averaged metric.
588 "precision": 3.14, # Precision is the fraction of actual positive predictions that had
589 # positive actual labels. For multiclass this is a macro-averaged
590 # metric treating each class as a binary classifier.
591 "logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.
592 "threshold": 3.14, # Threshold at which the metrics are computed. For binary
593 # classification models this is the positive class threshold.
594 # For multi-class classfication models this is the confidence
595 # threshold.
596 "accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For
597 # multiclass this is a micro-averaged metric.
598 "f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass
599 # this is a macro-averaged metric.
600 "rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged
601 # metric.
602 },
603 "positiveLabel": "A String", # Label representing the positive class.
604 "binaryConfusionMatrixList": [ # Binary confusion matrix at multiple thresholds.
605 { # Confusion matrix for binary classification models.
606 "truePositives": "A String", # Number of true samples predicted as true.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]607 "recall": 3.14, # The fraction of actual positive labels that were given a positive
608 # prediction.
609 "precision": 3.14, # The fraction of actual positive predictions that had positive actual
610 # labels.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]611 "falseNegatives": "A String", # Number of false samples predicted as false.
612 "trueNegatives": "A String", # Number of true samples predicted as false.
613 "falsePositives": "A String", # Number of false samples predicted as true.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]614 "f1Score": 3.14, # The equally weighted average of recall and precision.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]615 "positiveClassThreshold": 3.14, # Threshold value used when computing each of the following metric.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]616 "accuracy": 3.14, # The fraction of predictions given the correct label.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]617 },
618 ],
619 },
620 "multiClassClassificationMetrics": { # Evaluation metrics for multi-class classification/classifier models. # Populated for multi-class classification/classifier models.
621 "aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.
622 # models, the metrics are either macro-averaged or micro-averaged. When
623 # macro-averaged, the metrics are calculated for each label and then an
624 # unweighted average is taken of those values. When micro-averaged, the
625 # metric is calculated globally by counting the total number of correctly
626 # predicted rows.
627 "recall": 3.14, # Recall is the fraction of actual positive labels that were given a
628 # positive prediction. For multiclass this is a macro-averaged metric.
629 "precision": 3.14, # Precision is the fraction of actual positive predictions that had
630 # positive actual labels. For multiclass this is a macro-averaged
631 # metric treating each class as a binary classifier.
632 "logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.
633 "threshold": 3.14, # Threshold at which the metrics are computed. For binary
634 # classification models this is the positive class threshold.
635 # For multi-class classfication models this is the confidence
636 # threshold.
637 "accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For
638 # multiclass this is a micro-averaged metric.
639 "f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass
640 # this is a macro-averaged metric.
641 "rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged
642 # metric.
643 },
644 "confusionMatrixList": [ # Confusion matrix at different thresholds.
645 { # Confusion matrix for multi-class classification models.
646 "confidenceThreshold": 3.14, # Confidence threshold used when computing the entries of the
647 # confusion matrix.
648 "rows": [ # One row per actual label.
649 { # A single row in the confusion matrix.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]650 "actualLabel": "A String", # The original label of this row.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]651 "entries": [ # Info describing predicted label distribution.
652 { # A single entry in the confusion matrix.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]653 "predictedLabel": "A String", # The predicted label. For confidence_threshold &gt; 0, we will
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]654 # also add an entry indicating the number of items under the
655 # confidence threshold.
656 "itemCount": "A String", # Number of items being predicted as this label.
657 },
658 ],
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]659 },
660 ],
661 },
662 ],
663 },
664 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]665 "dataSplitResult": { # Data split result. This contains references to the training and evaluation # Data split result of the training run. Only set when the input data is
666 # actually split.
667 # data tables that were used to train the model.
668 "trainingTable": { # Table reference of the training data after split.
669 "projectId": "A String", # [Required] The ID of the project containing this table.
670 "tableId": "A String", # [Required] The ID of the table. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
671 "datasetId": "A String", # [Required] The ID of the dataset containing this table.
672 },
673 "evaluationTable": { # Table reference of the evaluation data after split.
674 "projectId": "A String", # [Required] The ID of the project containing this table.
675 "tableId": "A String", # [Required] The ID of the table. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
676 "datasetId": "A String", # [Required] The ID of the dataset containing this table.
677 },
678 },
679 "results": [ # Output of each iteration run, results.size() &lt;= max_iterations.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]680 { # Information about a single iteration of the training run.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]681 "arimaResult": { # (Auto-)arima fitting result. Wrap everything in ArimaResult for easier
682 # refactoring if we want to use model-specific iteration results.
683 "arimaModelInfo": [ # This message is repeated because there are multiple arima models
684 # fitted in auto-arima. For non-auto-arima model, its size is one.
685 { # Arima model information.
686 "seasonalPeriods": [ # Seasonal periods. Repeated because multiple periods are supported
687 # for one time series.
688 "A String",
689 ],
690 "hasDrift": True or False, # Whether Arima model fitted with drift or not. It is always false
691 # when d is not 1.
692 "arimaCoefficients": { # Arima coefficients. # Arima coefficients.
693 "movingAverageCoefficients": [ # Moving-average coefficients, an array of double.
694 3.14,
695 ],
696 "autoRegressiveCoefficients": [ # Auto-regressive coefficients, an array of double.
697 3.14,
698 ],
699 "interceptCoefficient": 3.14, # Intercept coefficient, just a double not an array.
700 },
701 "nonSeasonalOrder": { # Arima order, can be used for both non-seasonal and seasonal parts. # Non-seasonal order.
702 "q": "A String", # Order of the moving-average part.
703 "p": "A String", # Order of the autoregressive part.
704 "d": "A String", # Order of the differencing part.
705 },
706 "arimaFittingMetrics": { # ARIMA model fitting metrics. # Arima fitting metrics.
707 "variance": 3.14, # Variance.
708 "logLikelihood": 3.14, # Log-likelihood.
709 "aic": 3.14, # AIC.
710 },
711 "timeSeriesId": "A String", # The id to indicate different time series.
712 },
713 ],
714 "seasonalPeriods": [ # Seasonal periods. Repeated because multiple periods are supported for
715 # one time series.
716 "A String",
717 ],
718 },
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]719 "index": 42, # Index of the iteration, 0 based.
720 "evalLoss": 3.14, # Loss computed on the eval data at the end of iteration.
721 "durationMs": "A String", # Time taken to run the iteration in milliseconds.
722 "learnRate": 3.14, # Learn rate used for this iteration.
723 "trainingLoss": 3.14, # Loss computed on the training data at the end of iteration.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]724 "clusterInfos": [ # Information about top clusters for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]725 { # Information about a single cluster for clustering model.
726 "centroidId": "A String", # Centroid id.
727 "clusterSize": "A String", # Cluster size, the total number of points assigned to the cluster.
728 "clusterRadius": 3.14, # Cluster radius, the average distance from centroid
729 # to each point assigned to the cluster.
730 },
731 ],
732 },
733 ],
734 "startTime": "A String", # The start time of this training run.
735 "trainingOptions": { # Options that were used for this training run, includes
736 # user specified and default options that were used.
737 "optimizationStrategy": "A String", # Optimization strategy for training linear regression models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]738 "itemColumn": "A String", # Item column specified for matrix factorization models.
739 "feedbackType": "A String", # Feedback type that specifies which algorithm to run for matrix
740 # factorization.
741 "numFactors": "A String", # Num factors specified for matrix factorization models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]742 "inputLabelColumns": [ # Name of input label columns in training data.
743 "A String",
744 ],
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]745 "batchSize": "A String", # Batch size for dnn models.
746 "distanceType": "A String", # Distance type for clustering models.
747 "kmeansInitializationColumn": "A String", # The column used to provide the initial centroids for kmeans algorithm
748 # when kmeans_initialization_method is CUSTOM.
749 "l2Regularization": 3.14, # L2 regularization coefficient.
750 "dropout": 3.14, # Dropout probability for dnn models.
751 "minRelativeProgress": 3.14, # When early_stop is true, stops training when accuracy improvement is
752 # less than 'min_relative_progress'. Used only for iterative training
753 # algorithms.
754 "l1Regularization": 3.14, # L1 regularization coefficient.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]755 "maxIterations": "A String", # The maximum number of iterations in training. Used only for iterative
756 # training algorithms.
757 "earlyStop": True or False, # Whether to stop early when the loss doesn't improve significantly
758 # any more (compared to min_relative_progress). Used only for iterative
759 # training algorithms.
760 "initialLearnRate": 3.14, # Specifies the initial learning rate for the line search learn rate
761 # strategy.
762 "dataSplitColumn": "A String", # The column to split data with. This column won't be used as a
763 # feature.
764 # 1. When data_split_method is CUSTOM, the corresponding column should
765 # be boolean. The rows with true value tag are eval data, and the false
766 # are training data.
767 # 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION
768 # rows (from smallest to largest) in the corresponding column are used
769 # as training data, and the rest are eval data. It respects the order
770 # in Orderable data types:
771 # https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data-type-properties
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]772 "numClusters": "A String", # Number of clusters for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]773 "warmStart": True or False, # Whether to train a model from the last checkpoint.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]774 "hiddenUnits": [ # Hidden units for dnn models.
775 "A String",
776 ],
777 "maxTreeDepth": "A String", # Maximum depth of a tree for boosted tree models.
778 "userColumn": "A String", # User column specified for matrix factorization models.
779 "kmeansInitializationMethod": "A String", # The method used to initialize the centroids for kmeans algorithm.
780 "learnRateStrategy": "A String", # The strategy to determine learn rate for the current iteration.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]781 "dataSplitEvalFraction": 3.14, # The fraction of evaluation data over the whole input data. The rest
782 # of data will be used as training data. The format should be double.
783 # Accurate to two decimal places.
784 # Default value is 0.2.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]785 "dataSplitMethod": "A String", # The data split type for training and evaluation, e.g. RANDOM.
786 "subsample": 3.14, # Subsample fraction of the training data to grow tree to prevent
787 # overfitting for boosted tree models.
788 "labelClassWeights": { # Weights associated with each label class, for rebalancing the
789 # training data. Only applicable for classification models.
790 "a_key": 3.14,
791 },
792 "learnRate": 3.14, # Learning rate in training. Used only for iterative training algorithms.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]793 "modelUri": "A String", # [Beta] Google Cloud Storage URI from which the model was imported. Only
794 # applicable for imported models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]795 "walsAlpha": 3.14, # Hyperparameter for matrix factoration when implicit feedback type is
796 # specified.
797 "minSplitLoss": 3.14, # Minimum split loss for boosted tree models.
798 "lossType": "A String", # Type of loss function used during training run.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]799 },
800 },
801 ],
802 "featureColumns": [ # Output only. Input feature columns that were used to train this model.
803 { # A field or a column.
804 "type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly
805 # specified (e.g., CREATE FUNCTION statement can omit the return type;
806 # in this case the output parameter does not have this "type" field).
807 # Examples:
808 # INT64: {type_kind="INT64"}
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]809 # ARRAY&lt;STRING&gt;: {type_kind="ARRAY", array_element_type="STRING"}
810 # STRUCT&lt;x STRING, y ARRAY&lt;DATE&gt;&gt;:
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]811 # {type_kind="STRUCT",
812 # struct_type={fields=[
813 # {name="x", type={type_kind="STRING"}},
814 # {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}
815 # ]}}
816 "structType": { # The fields of this struct, in order, if type_kind = "STRUCT".
817 "fields": [
818 # Object with schema name: StandardSqlField
819 ],
820 },
821 "arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".
822 "typeKind": "A String", # Required. The top level type of this field.
823 # Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").
824 },
825 "name": "A String", # Optional. The name of this field. Can be absent for struct fields.
826 },
827 ],
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]828 "labelColumns": [ # Output only. Label columns that were used to train this model.
829 # The output of the model will have a "predicted_" prefix to these columns.
830 { # A field or a column.
831 "type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly
832 # specified (e.g., CREATE FUNCTION statement can omit the return type;
833 # in this case the output parameter does not have this "type" field).
834 # Examples:
835 # INT64: {type_kind="INT64"}
836 # ARRAY&lt;STRING&gt;: {type_kind="ARRAY", array_element_type="STRING"}
837 # STRUCT&lt;x STRING, y ARRAY&lt;DATE&gt;&gt;:
838 # {type_kind="STRUCT",
839 # struct_type={fields=[
840 # {name="x", type={type_kind="STRING"}},
841 # {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}
842 # ]}}
843 "structType": { # The fields of this struct, in order, if type_kind = "STRUCT".
844 "fields": [
845 # Object with schema name: StandardSqlField
846 ],
847 },
848 "arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".
849 "typeKind": "A String", # Required. The top level type of this field.
850 # Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").
851 },
852 "name": "A String", # Optional. The name of this field. Can be absent for struct fields.
853 },
854 ],
855 "creationTime": "A String", # Output only. The time when this model was created, in millisecs since the epoch.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]856 "modelType": "A String", # Output only. Type of the model resource.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]857 "encryptionConfiguration": { # Custom encryption configuration (e.g., Cloud KMS keys). This shows the
858 # encryption configuration of the model data while stored in BigQuery
859 # storage. This field can be used with PatchModel to update encryption key
860 # for an already encrypted model.
861 "kmsKeyName": "A String", # [Optional] Describes the Cloud KMS encryption key that will be used to protect destination BigQuery table. The BigQuery Service Account associated with your project requires access to this encryption key.
862 },
863 "modelReference": { # Required. Unique identifier for this model.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]864 "projectId": "A String", # [Required] The ID of the project containing this model.
865 "datasetId": "A String", # [Required] The ID of the dataset containing this model.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]866 "modelId": "A String", # [Required] The ID of the model. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]867 },
868 "etag": "A String", # Output only. A hash of this resource.
869 "location": "A String", # Output only. The geographic location where the model resides. This value
870 # is inherited from the dataset.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]871 "friendlyName": "A String", # Optional. A descriptive name for this model.
872 "expirationTime": "A String", # Optional. The time when this model expires, in milliseconds since the epoch.
873 # If not present, the model will persist indefinitely. Expired models
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]874 # will be deleted and their storage reclaimed. The defaultTableExpirationMs
875 # property of the encapsulating dataset can be used to set a default
876 # expirationTime on newly created models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]877 "lastModifiedTime": "A String", # Output only. The time when this model was last modified, in millisecs since the epoch.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]878 },
879 ],
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]880 }</pre>
881</div>
882
883<div class="method">
884 <code class="details" id="list_next">list_next(previous_request, previous_response)</code>
885 <pre>Retrieves the next page of results.
886
887Args:
888 previous_request: The request for the previous page. (required)
889 previous_response: The response from the request for the previous page. (required)
890
891Returns:
892 A request object that you can call 'execute()' on to request the next
893 page. Returns None if there are no more items in the collection.
894 </pre>
895</div>
896
897<div class="method">
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]898 <code class="details" id="patch">patch(projectId, datasetId, modelId, body=None)</code>
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]899 <pre>Patch specific fields in the specified model.
900
901Args:
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]902 projectId: string, Required. Project ID of the model to patch. (required)
903 datasetId: string, Required. Dataset ID of the model to patch. (required)
904 modelId: string, Required. Model ID of the model to patch. (required)
905 body: object, The request body.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]906 The object takes the form of:
907
908{
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]909 "labels": { # The labels associated with this model. You can use these to organize
910 # and group your models. Label keys and values can be no longer
911 # than 63 characters, can only contain lowercase letters, numeric
912 # characters, underscores and dashes. International characters are allowed.
913 # Label values are optional. Label keys must start with a letter and each
914 # label in the list must have a different key.
915 "a_key": "A String",
916 },
917 "description": "A String", # Optional. A user-friendly description of this model.
918 "trainingRuns": [ # Output only. Information for all training runs in increasing order of start_time.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]919 { # Information about a single training query run for the model.
920 "evaluationMetrics": { # Evaluation metrics of a model. These are either computed on all training # The evaluation metrics over training/eval data that were computed at the
921 # end of training.
922 # data or just the eval data based on whether eval data was used during
923 # training. These are not present for imported models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]924 "clusteringMetrics": { # Evaluation metrics for clustering models. # Populated for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]925 "meanSquaredDistance": 3.14, # Mean of squared distances between each sample to its cluster centroid.
926 "daviesBouldinIndex": 3.14, # Davies-Bouldin index.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]927 "clusters": [ # [Beta] Information for all clusters.
928 { # Message containing the information about one cluster.
929 "count": "A String", # Count of training data rows that were assigned to this cluster.
930 "featureValues": [ # Values of highly variant features for this cluster.
931 { # Representative value of a single feature within the cluster.
932 "featureColumn": "A String", # The feature column name.
933 "numericalValue": 3.14, # The numerical feature value. This is the centroid value for this
934 # feature.
935 "categoricalValue": { # Representative value of a categorical feature. # The categorical feature value.
936 "categoryCounts": [ # Counts of all categories for the categorical feature. If there are
937 # more than ten categories, we return top ten (by count) and return
938 # one more CategoryCount with category "_OTHER_" and count as
939 # aggregate counts of remaining categories.
940 { # Represents the count of a single category within the cluster.
941 "category": "A String", # The name of category.
942 "count": "A String", # The count of training samples matching the category within the
943 # cluster.
944 },
945 ],
946 },
947 },
948 ],
949 "centroidId": "A String", # Centroid id.
950 },
951 ],
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]952 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]953 "regressionMetrics": { # Evaluation metrics for regression and explicit feedback type matrix # Populated for regression models and explicit feedback type matrix
954 # factorization models.
955 # factorization models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]956 "meanSquaredLogError": 3.14, # Mean squared log error.
957 "meanAbsoluteError": 3.14, # Mean absolute error.
958 "meanSquaredError": 3.14, # Mean squared error.
959 "medianAbsoluteError": 3.14, # Median absolute error.
960 "rSquared": 3.14, # R^2 score.
961 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]962 "rankingMetrics": { # Evaluation metrics used by weighted-ALS models specified by # [Alpha] Populated for implicit feedback type matrix factorization
963 # models.
964 # feedback_type=implicit.
965 "meanSquaredError": 3.14, # Similar to the mean squared error computed in regression and explicit
966 # recommendation models except instead of computing the rating directly,
967 # the output from evaluate is computed against a preference which is 1 or 0
968 # depending on if the rating exists or not.
969 "meanAveragePrecision": 3.14, # Calculates a precision per user for all the items by ranking them and
970 # then averages all the precisions across all the users.
971 "averageRank": 3.14, # Determines the goodness of a ranking by computing the percentile rank
972 # from the predicted confidence and dividing it by the original rank.
973 "normalizedDiscountedCumulativeGain": 3.14, # A metric to determine the goodness of a ranking calculated from the
974 # predicted confidence by comparing it to an ideal rank measured by the
975 # original ratings.
976 },
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]977 "binaryClassificationMetrics": { # Evaluation metrics for binary classification/classifier models. # Populated for binary classification/classifier models.
978 "negativeLabel": "A String", # Label representing the negative class.
979 "aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.
980 # models, the metrics are either macro-averaged or micro-averaged. When
981 # macro-averaged, the metrics are calculated for each label and then an
982 # unweighted average is taken of those values. When micro-averaged, the
983 # metric is calculated globally by counting the total number of correctly
984 # predicted rows.
985 "recall": 3.14, # Recall is the fraction of actual positive labels that were given a
986 # positive prediction. For multiclass this is a macro-averaged metric.
987 "precision": 3.14, # Precision is the fraction of actual positive predictions that had
988 # positive actual labels. For multiclass this is a macro-averaged
989 # metric treating each class as a binary classifier.
990 "logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.
991 "threshold": 3.14, # Threshold at which the metrics are computed. For binary
992 # classification models this is the positive class threshold.
993 # For multi-class classfication models this is the confidence
994 # threshold.
995 "accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For
996 # multiclass this is a micro-averaged metric.
997 "f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass
998 # this is a macro-averaged metric.
999 "rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged
1000 # metric.
1001 },
1002 "positiveLabel": "A String", # Label representing the positive class.
1003 "binaryConfusionMatrixList": [ # Binary confusion matrix at multiple thresholds.
1004 { # Confusion matrix for binary classification models.
1005 "truePositives": "A String", # Number of true samples predicted as true.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1006 "recall": 3.14, # The fraction of actual positive labels that were given a positive
1007 # prediction.
1008 "precision": 3.14, # The fraction of actual positive predictions that had positive actual
1009 # labels.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1010 "falseNegatives": "A String", # Number of false samples predicted as false.
1011 "trueNegatives": "A String", # Number of true samples predicted as false.
1012 "falsePositives": "A String", # Number of false samples predicted as true.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1013 "f1Score": 3.14, # The equally weighted average of recall and precision.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1014 "positiveClassThreshold": 3.14, # Threshold value used when computing each of the following metric.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1015 "accuracy": 3.14, # The fraction of predictions given the correct label.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1016 },
1017 ],
1018 },
1019 "multiClassClassificationMetrics": { # Evaluation metrics for multi-class classification/classifier models. # Populated for multi-class classification/classifier models.
1020 "aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.
1021 # models, the metrics are either macro-averaged or micro-averaged. When
1022 # macro-averaged, the metrics are calculated for each label and then an
1023 # unweighted average is taken of those values. When micro-averaged, the
1024 # metric is calculated globally by counting the total number of correctly
1025 # predicted rows.
1026 "recall": 3.14, # Recall is the fraction of actual positive labels that were given a
1027 # positive prediction. For multiclass this is a macro-averaged metric.
1028 "precision": 3.14, # Precision is the fraction of actual positive predictions that had
1029 # positive actual labels. For multiclass this is a macro-averaged
1030 # metric treating each class as a binary classifier.
1031 "logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.
1032 "threshold": 3.14, # Threshold at which the metrics are computed. For binary
1033 # classification models this is the positive class threshold.
1034 # For multi-class classfication models this is the confidence
1035 # threshold.
1036 "accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For
1037 # multiclass this is a micro-averaged metric.
1038 "f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass
1039 # this is a macro-averaged metric.
1040 "rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged
1041 # metric.
1042 },
1043 "confusionMatrixList": [ # Confusion matrix at different thresholds.
1044 { # Confusion matrix for multi-class classification models.
1045 "confidenceThreshold": 3.14, # Confidence threshold used when computing the entries of the
1046 # confusion matrix.
1047 "rows": [ # One row per actual label.
1048 { # A single row in the confusion matrix.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1049 "actualLabel": "A String", # The original label of this row.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1050 "entries": [ # Info describing predicted label distribution.
1051 { # A single entry in the confusion matrix.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1052 "predictedLabel": "A String", # The predicted label. For confidence_threshold &gt; 0, we will
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1053 # also add an entry indicating the number of items under the
1054 # confidence threshold.
1055 "itemCount": "A String", # Number of items being predicted as this label.
1056 },
1057 ],
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1058 },
1059 ],
1060 },
1061 ],
1062 },
1063 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1064 "dataSplitResult": { # Data split result. This contains references to the training and evaluation # Data split result of the training run. Only set when the input data is
1065 # actually split.
1066 # data tables that were used to train the model.
1067 "trainingTable": { # Table reference of the training data after split.
1068 "projectId": "A String", # [Required] The ID of the project containing this table.
1069 "tableId": "A String", # [Required] The ID of the table. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
1070 "datasetId": "A String", # [Required] The ID of the dataset containing this table.
1071 },
1072 "evaluationTable": { # Table reference of the evaluation data after split.
1073 "projectId": "A String", # [Required] The ID of the project containing this table.
1074 "tableId": "A String", # [Required] The ID of the table. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
1075 "datasetId": "A String", # [Required] The ID of the dataset containing this table.
1076 },
1077 },
1078 "results": [ # Output of each iteration run, results.size() &lt;= max_iterations.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1079 { # Information about a single iteration of the training run.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1080 "arimaResult": { # (Auto-)arima fitting result. Wrap everything in ArimaResult for easier
1081 # refactoring if we want to use model-specific iteration results.
1082 "arimaModelInfo": [ # This message is repeated because there are multiple arima models
1083 # fitted in auto-arima. For non-auto-arima model, its size is one.
1084 { # Arima model information.
1085 "seasonalPeriods": [ # Seasonal periods. Repeated because multiple periods are supported
1086 # for one time series.
1087 "A String",
1088 ],
1089 "hasDrift": True or False, # Whether Arima model fitted with drift or not. It is always false
1090 # when d is not 1.
1091 "arimaCoefficients": { # Arima coefficients. # Arima coefficients.
1092 "movingAverageCoefficients": [ # Moving-average coefficients, an array of double.
1093 3.14,
1094 ],
1095 "autoRegressiveCoefficients": [ # Auto-regressive coefficients, an array of double.
1096 3.14,
1097 ],
1098 "interceptCoefficient": 3.14, # Intercept coefficient, just a double not an array.
1099 },
1100 "nonSeasonalOrder": { # Arima order, can be used for both non-seasonal and seasonal parts. # Non-seasonal order.
1101 "q": "A String", # Order of the moving-average part.
1102 "p": "A String", # Order of the autoregressive part.
1103 "d": "A String", # Order of the differencing part.
1104 },
1105 "arimaFittingMetrics": { # ARIMA model fitting metrics. # Arima fitting metrics.
1106 "variance": 3.14, # Variance.
1107 "logLikelihood": 3.14, # Log-likelihood.
1108 "aic": 3.14, # AIC.
1109 },
1110 "timeSeriesId": "A String", # The id to indicate different time series.
1111 },
1112 ],
1113 "seasonalPeriods": [ # Seasonal periods. Repeated because multiple periods are supported for
1114 # one time series.
1115 "A String",
1116 ],
1117 },
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1118 "index": 42, # Index of the iteration, 0 based.
1119 "evalLoss": 3.14, # Loss computed on the eval data at the end of iteration.
1120 "durationMs": "A String", # Time taken to run the iteration in milliseconds.
1121 "learnRate": 3.14, # Learn rate used for this iteration.
1122 "trainingLoss": 3.14, # Loss computed on the training data at the end of iteration.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1123 "clusterInfos": [ # Information about top clusters for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1124 { # Information about a single cluster for clustering model.
1125 "centroidId": "A String", # Centroid id.
1126 "clusterSize": "A String", # Cluster size, the total number of points assigned to the cluster.
1127 "clusterRadius": 3.14, # Cluster radius, the average distance from centroid
1128 # to each point assigned to the cluster.
1129 },
1130 ],
1131 },
1132 ],
1133 "startTime": "A String", # The start time of this training run.
1134 "trainingOptions": { # Options that were used for this training run, includes
1135 # user specified and default options that were used.
1136 "optimizationStrategy": "A String", # Optimization strategy for training linear regression models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1137 "itemColumn": "A String", # Item column specified for matrix factorization models.
1138 "feedbackType": "A String", # Feedback type that specifies which algorithm to run for matrix
1139 # factorization.
1140 "numFactors": "A String", # Num factors specified for matrix factorization models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1141 "inputLabelColumns": [ # Name of input label columns in training data.
1142 "A String",
1143 ],
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1144 "batchSize": "A String", # Batch size for dnn models.
1145 "distanceType": "A String", # Distance type for clustering models.
1146 "kmeansInitializationColumn": "A String", # The column used to provide the initial centroids for kmeans algorithm
1147 # when kmeans_initialization_method is CUSTOM.
1148 "l2Regularization": 3.14, # L2 regularization coefficient.
1149 "dropout": 3.14, # Dropout probability for dnn models.
1150 "minRelativeProgress": 3.14, # When early_stop is true, stops training when accuracy improvement is
1151 # less than 'min_relative_progress'. Used only for iterative training
1152 # algorithms.
1153 "l1Regularization": 3.14, # L1 regularization coefficient.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1154 "maxIterations": "A String", # The maximum number of iterations in training. Used only for iterative
1155 # training algorithms.
1156 "earlyStop": True or False, # Whether to stop early when the loss doesn't improve significantly
1157 # any more (compared to min_relative_progress). Used only for iterative
1158 # training algorithms.
1159 "initialLearnRate": 3.14, # Specifies the initial learning rate for the line search learn rate
1160 # strategy.
1161 "dataSplitColumn": "A String", # The column to split data with. This column won't be used as a
1162 # feature.
1163 # 1. When data_split_method is CUSTOM, the corresponding column should
1164 # be boolean. The rows with true value tag are eval data, and the false
1165 # are training data.
1166 # 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION
1167 # rows (from smallest to largest) in the corresponding column are used
1168 # as training data, and the rest are eval data. It respects the order
1169 # in Orderable data types:
1170 # https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data-type-properties
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1171 "numClusters": "A String", # Number of clusters for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1172 "warmStart": True or False, # Whether to train a model from the last checkpoint.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1173 "hiddenUnits": [ # Hidden units for dnn models.
1174 "A String",
1175 ],
1176 "maxTreeDepth": "A String", # Maximum depth of a tree for boosted tree models.
1177 "userColumn": "A String", # User column specified for matrix factorization models.
1178 "kmeansInitializationMethod": "A String", # The method used to initialize the centroids for kmeans algorithm.
1179 "learnRateStrategy": "A String", # The strategy to determine learn rate for the current iteration.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1180 "dataSplitEvalFraction": 3.14, # The fraction of evaluation data over the whole input data. The rest
1181 # of data will be used as training data. The format should be double.
1182 # Accurate to two decimal places.
1183 # Default value is 0.2.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1184 "dataSplitMethod": "A String", # The data split type for training and evaluation, e.g. RANDOM.
1185 "subsample": 3.14, # Subsample fraction of the training data to grow tree to prevent
1186 # overfitting for boosted tree models.
1187 "labelClassWeights": { # Weights associated with each label class, for rebalancing the
1188 # training data. Only applicable for classification models.
1189 "a_key": 3.14,
1190 },
1191 "learnRate": 3.14, # Learning rate in training. Used only for iterative training algorithms.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1192 "modelUri": "A String", # [Beta] Google Cloud Storage URI from which the model was imported. Only
1193 # applicable for imported models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1194 "walsAlpha": 3.14, # Hyperparameter for matrix factoration when implicit feedback type is
1195 # specified.
1196 "minSplitLoss": 3.14, # Minimum split loss for boosted tree models.
1197 "lossType": "A String", # Type of loss function used during training run.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1198 },
1199 },
1200 ],
1201 "featureColumns": [ # Output only. Input feature columns that were used to train this model.
1202 { # A field or a column.
1203 "type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly
1204 # specified (e.g., CREATE FUNCTION statement can omit the return type;
1205 # in this case the output parameter does not have this "type" field).
1206 # Examples:
1207 # INT64: {type_kind="INT64"}
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1208 # ARRAY&lt;STRING&gt;: {type_kind="ARRAY", array_element_type="STRING"}
1209 # STRUCT&lt;x STRING, y ARRAY&lt;DATE&gt;&gt;:
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1210 # {type_kind="STRUCT",
1211 # struct_type={fields=[
1212 # {name="x", type={type_kind="STRING"}},
1213 # {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}
1214 # ]}}
1215 "structType": { # The fields of this struct, in order, if type_kind = "STRUCT".
1216 "fields": [
1217 # Object with schema name: StandardSqlField
1218 ],
1219 },
1220 "arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".
1221 "typeKind": "A String", # Required. The top level type of this field.
1222 # Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").
1223 },
1224 "name": "A String", # Optional. The name of this field. Can be absent for struct fields.
1225 },
1226 ],
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1227 "labelColumns": [ # Output only. Label columns that were used to train this model.
1228 # The output of the model will have a "predicted_" prefix to these columns.
1229 { # A field or a column.
1230 "type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly
1231 # specified (e.g., CREATE FUNCTION statement can omit the return type;
1232 # in this case the output parameter does not have this "type" field).
1233 # Examples:
1234 # INT64: {type_kind="INT64"}
1235 # ARRAY&lt;STRING&gt;: {type_kind="ARRAY", array_element_type="STRING"}
1236 # STRUCT&lt;x STRING, y ARRAY&lt;DATE&gt;&gt;:
1237 # {type_kind="STRUCT",
1238 # struct_type={fields=[
1239 # {name="x", type={type_kind="STRING"}},
1240 # {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}
1241 # ]}}
1242 "structType": { # The fields of this struct, in order, if type_kind = "STRUCT".
1243 "fields": [
1244 # Object with schema name: StandardSqlField
1245 ],
1246 },
1247 "arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".
1248 "typeKind": "A String", # Required. The top level type of this field.
1249 # Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").
1250 },
1251 "name": "A String", # Optional. The name of this field. Can be absent for struct fields.
1252 },
1253 ],
1254 "creationTime": "A String", # Output only. The time when this model was created, in millisecs since the epoch.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1255 "modelType": "A String", # Output only. Type of the model resource.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1256 "encryptionConfiguration": { # Custom encryption configuration (e.g., Cloud KMS keys). This shows the
1257 # encryption configuration of the model data while stored in BigQuery
1258 # storage. This field can be used with PatchModel to update encryption key
1259 # for an already encrypted model.
1260 "kmsKeyName": "A String", # [Optional] Describes the Cloud KMS encryption key that will be used to protect destination BigQuery table. The BigQuery Service Account associated with your project requires access to this encryption key.
1261 },
1262 "modelReference": { # Required. Unique identifier for this model.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1263 "projectId": "A String", # [Required] The ID of the project containing this model.
1264 "datasetId": "A String", # [Required] The ID of the dataset containing this model.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1265 "modelId": "A String", # [Required] The ID of the model. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1266 },
1267 "etag": "A String", # Output only. A hash of this resource.
1268 "location": "A String", # Output only. The geographic location where the model resides. This value
1269 # is inherited from the dataset.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1270 "friendlyName": "A String", # Optional. A descriptive name for this model.
1271 "expirationTime": "A String", # Optional. The time when this model expires, in milliseconds since the epoch.
1272 # If not present, the model will persist indefinitely. Expired models
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1273 # will be deleted and their storage reclaimed. The defaultTableExpirationMs
1274 # property of the encapsulating dataset can be used to set a default
1275 # expirationTime on newly created models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1276 "lastModifiedTime": "A String", # Output only. The time when this model was last modified, in millisecs since the epoch.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1277 }
1278
1279
1280Returns:
1281 An object of the form:
1282
1283 {
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1284 "labels": { # The labels associated with this model. You can use these to organize
1285 # and group your models. Label keys and values can be no longer
1286 # than 63 characters, can only contain lowercase letters, numeric
1287 # characters, underscores and dashes. International characters are allowed.
1288 # Label values are optional. Label keys must start with a letter and each
1289 # label in the list must have a different key.
1290 "a_key": "A String",
1291 },
1292 "description": "A String", # Optional. A user-friendly description of this model.
1293 "trainingRuns": [ # Output only. Information for all training runs in increasing order of start_time.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1294 { # Information about a single training query run for the model.
1295 "evaluationMetrics": { # Evaluation metrics of a model. These are either computed on all training # The evaluation metrics over training/eval data that were computed at the
1296 # end of training.
1297 # data or just the eval data based on whether eval data was used during
1298 # training. These are not present for imported models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1299 "clusteringMetrics": { # Evaluation metrics for clustering models. # Populated for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1300 "meanSquaredDistance": 3.14, # Mean of squared distances between each sample to its cluster centroid.
1301 "daviesBouldinIndex": 3.14, # Davies-Bouldin index.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1302 "clusters": [ # [Beta] Information for all clusters.
1303 { # Message containing the information about one cluster.
1304 "count": "A String", # Count of training data rows that were assigned to this cluster.
1305 "featureValues": [ # Values of highly variant features for this cluster.
1306 { # Representative value of a single feature within the cluster.
1307 "featureColumn": "A String", # The feature column name.
1308 "numericalValue": 3.14, # The numerical feature value. This is the centroid value for this
1309 # feature.
1310 "categoricalValue": { # Representative value of a categorical feature. # The categorical feature value.
1311 "categoryCounts": [ # Counts of all categories for the categorical feature. If there are
1312 # more than ten categories, we return top ten (by count) and return
1313 # one more CategoryCount with category "_OTHER_" and count as
1314 # aggregate counts of remaining categories.
1315 { # Represents the count of a single category within the cluster.
1316 "category": "A String", # The name of category.
1317 "count": "A String", # The count of training samples matching the category within the
1318 # cluster.
1319 },
1320 ],
1321 },
1322 },
1323 ],
1324 "centroidId": "A String", # Centroid id.
1325 },
1326 ],
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1327 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1328 "regressionMetrics": { # Evaluation metrics for regression and explicit feedback type matrix # Populated for regression models and explicit feedback type matrix
1329 # factorization models.
1330 # factorization models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1331 "meanSquaredLogError": 3.14, # Mean squared log error.
1332 "meanAbsoluteError": 3.14, # Mean absolute error.
1333 "meanSquaredError": 3.14, # Mean squared error.
1334 "medianAbsoluteError": 3.14, # Median absolute error.
1335 "rSquared": 3.14, # R^2 score.
1336 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1337 "rankingMetrics": { # Evaluation metrics used by weighted-ALS models specified by # [Alpha] Populated for implicit feedback type matrix factorization
1338 # models.
1339 # feedback_type=implicit.
1340 "meanSquaredError": 3.14, # Similar to the mean squared error computed in regression and explicit
1341 # recommendation models except instead of computing the rating directly,
1342 # the output from evaluate is computed against a preference which is 1 or 0
1343 # depending on if the rating exists or not.
1344 "meanAveragePrecision": 3.14, # Calculates a precision per user for all the items by ranking them and
1345 # then averages all the precisions across all the users.
1346 "averageRank": 3.14, # Determines the goodness of a ranking by computing the percentile rank
1347 # from the predicted confidence and dividing it by the original rank.
1348 "normalizedDiscountedCumulativeGain": 3.14, # A metric to determine the goodness of a ranking calculated from the
1349 # predicted confidence by comparing it to an ideal rank measured by the
1350 # original ratings.
1351 },
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1352 "binaryClassificationMetrics": { # Evaluation metrics for binary classification/classifier models. # Populated for binary classification/classifier models.
1353 "negativeLabel": "A String", # Label representing the negative class.
1354 "aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.
1355 # models, the metrics are either macro-averaged or micro-averaged. When
1356 # macro-averaged, the metrics are calculated for each label and then an
1357 # unweighted average is taken of those values. When micro-averaged, the
1358 # metric is calculated globally by counting the total number of correctly
1359 # predicted rows.
1360 "recall": 3.14, # Recall is the fraction of actual positive labels that were given a
1361 # positive prediction. For multiclass this is a macro-averaged metric.
1362 "precision": 3.14, # Precision is the fraction of actual positive predictions that had
1363 # positive actual labels. For multiclass this is a macro-averaged
1364 # metric treating each class as a binary classifier.
1365 "logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.
1366 "threshold": 3.14, # Threshold at which the metrics are computed. For binary
1367 # classification models this is the positive class threshold.
1368 # For multi-class classfication models this is the confidence
1369 # threshold.
1370 "accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For
1371 # multiclass this is a micro-averaged metric.
1372 "f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass
1373 # this is a macro-averaged metric.
1374 "rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged
1375 # metric.
1376 },
1377 "positiveLabel": "A String", # Label representing the positive class.
1378 "binaryConfusionMatrixList": [ # Binary confusion matrix at multiple thresholds.
1379 { # Confusion matrix for binary classification models.
1380 "truePositives": "A String", # Number of true samples predicted as true.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1381 "recall": 3.14, # The fraction of actual positive labels that were given a positive
1382 # prediction.
1383 "precision": 3.14, # The fraction of actual positive predictions that had positive actual
1384 # labels.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1385 "falseNegatives": "A String", # Number of false samples predicted as false.
1386 "trueNegatives": "A String", # Number of true samples predicted as false.
1387 "falsePositives": "A String", # Number of false samples predicted as true.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1388 "f1Score": 3.14, # The equally weighted average of recall and precision.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1389 "positiveClassThreshold": 3.14, # Threshold value used when computing each of the following metric.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1390 "accuracy": 3.14, # The fraction of predictions given the correct label.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1391 },
1392 ],
1393 },
1394 "multiClassClassificationMetrics": { # Evaluation metrics for multi-class classification/classifier models. # Populated for multi-class classification/classifier models.
1395 "aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.
1396 # models, the metrics are either macro-averaged or micro-averaged. When
1397 # macro-averaged, the metrics are calculated for each label and then an
1398 # unweighted average is taken of those values. When micro-averaged, the
1399 # metric is calculated globally by counting the total number of correctly
1400 # predicted rows.
1401 "recall": 3.14, # Recall is the fraction of actual positive labels that were given a
1402 # positive prediction. For multiclass this is a macro-averaged metric.
1403 "precision": 3.14, # Precision is the fraction of actual positive predictions that had
1404 # positive actual labels. For multiclass this is a macro-averaged
1405 # metric treating each class as a binary classifier.
1406 "logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.
1407 "threshold": 3.14, # Threshold at which the metrics are computed. For binary
1408 # classification models this is the positive class threshold.
1409 # For multi-class classfication models this is the confidence
1410 # threshold.
1411 "accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For
1412 # multiclass this is a micro-averaged metric.
1413 "f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass
1414 # this is a macro-averaged metric.
1415 "rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged
1416 # metric.
1417 },
1418 "confusionMatrixList": [ # Confusion matrix at different thresholds.
1419 { # Confusion matrix for multi-class classification models.
1420 "confidenceThreshold": 3.14, # Confidence threshold used when computing the entries of the
1421 # confusion matrix.
1422 "rows": [ # One row per actual label.
1423 { # A single row in the confusion matrix.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1424 "actualLabel": "A String", # The original label of this row.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1425 "entries": [ # Info describing predicted label distribution.
1426 { # A single entry in the confusion matrix.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1427 "predictedLabel": "A String", # The predicted label. For confidence_threshold &gt; 0, we will
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1428 # also add an entry indicating the number of items under the
1429 # confidence threshold.
1430 "itemCount": "A String", # Number of items being predicted as this label.
1431 },
1432 ],
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1433 },
1434 ],
1435 },
1436 ],
1437 },
1438 },
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1439 "dataSplitResult": { # Data split result. This contains references to the training and evaluation # Data split result of the training run. Only set when the input data is
1440 # actually split.
1441 # data tables that were used to train the model.
1442 "trainingTable": { # Table reference of the training data after split.
1443 "projectId": "A String", # [Required] The ID of the project containing this table.
1444 "tableId": "A String", # [Required] The ID of the table. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
1445 "datasetId": "A String", # [Required] The ID of the dataset containing this table.
1446 },
1447 "evaluationTable": { # Table reference of the evaluation data after split.
1448 "projectId": "A String", # [Required] The ID of the project containing this table.
1449 "tableId": "A String", # [Required] The ID of the table. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
1450 "datasetId": "A String", # [Required] The ID of the dataset containing this table.
1451 },
1452 },
1453 "results": [ # Output of each iteration run, results.size() &lt;= max_iterations.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1454 { # Information about a single iteration of the training run.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1455 "arimaResult": { # (Auto-)arima fitting result. Wrap everything in ArimaResult for easier
1456 # refactoring if we want to use model-specific iteration results.
1457 "arimaModelInfo": [ # This message is repeated because there are multiple arima models
1458 # fitted in auto-arima. For non-auto-arima model, its size is one.
1459 { # Arima model information.
1460 "seasonalPeriods": [ # Seasonal periods. Repeated because multiple periods are supported
1461 # for one time series.
1462 "A String",
1463 ],
1464 "hasDrift": True or False, # Whether Arima model fitted with drift or not. It is always false
1465 # when d is not 1.
1466 "arimaCoefficients": { # Arima coefficients. # Arima coefficients.
1467 "movingAverageCoefficients": [ # Moving-average coefficients, an array of double.
1468 3.14,
1469 ],
1470 "autoRegressiveCoefficients": [ # Auto-regressive coefficients, an array of double.
1471 3.14,
1472 ],
1473 "interceptCoefficient": 3.14, # Intercept coefficient, just a double not an array.
1474 },
1475 "nonSeasonalOrder": { # Arima order, can be used for both non-seasonal and seasonal parts. # Non-seasonal order.
1476 "q": "A String", # Order of the moving-average part.
1477 "p": "A String", # Order of the autoregressive part.
1478 "d": "A String", # Order of the differencing part.
1479 },
1480 "arimaFittingMetrics": { # ARIMA model fitting metrics. # Arima fitting metrics.
1481 "variance": 3.14, # Variance.
1482 "logLikelihood": 3.14, # Log-likelihood.
1483 "aic": 3.14, # AIC.
1484 },
1485 "timeSeriesId": "A String", # The id to indicate different time series.
1486 },
1487 ],
1488 "seasonalPeriods": [ # Seasonal periods. Repeated because multiple periods are supported for
1489 # one time series.
1490 "A String",
1491 ],
1492 },
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1493 "index": 42, # Index of the iteration, 0 based.
1494 "evalLoss": 3.14, # Loss computed on the eval data at the end of iteration.
1495 "durationMs": "A String", # Time taken to run the iteration in milliseconds.
1496 "learnRate": 3.14, # Learn rate used for this iteration.
1497 "trainingLoss": 3.14, # Loss computed on the training data at the end of iteration.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1498 "clusterInfos": [ # Information about top clusters for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1499 { # Information about a single cluster for clustering model.
1500 "centroidId": "A String", # Centroid id.
1501 "clusterSize": "A String", # Cluster size, the total number of points assigned to the cluster.
1502 "clusterRadius": 3.14, # Cluster radius, the average distance from centroid
1503 # to each point assigned to the cluster.
1504 },
1505 ],
1506 },
1507 ],
1508 "startTime": "A String", # The start time of this training run.
1509 "trainingOptions": { # Options that were used for this training run, includes
1510 # user specified and default options that were used.
1511 "optimizationStrategy": "A String", # Optimization strategy for training linear regression models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1512 "itemColumn": "A String", # Item column specified for matrix factorization models.
1513 "feedbackType": "A String", # Feedback type that specifies which algorithm to run for matrix
1514 # factorization.
1515 "numFactors": "A String", # Num factors specified for matrix factorization models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1516 "inputLabelColumns": [ # Name of input label columns in training data.
1517 "A String",
1518 ],
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1519 "batchSize": "A String", # Batch size for dnn models.
1520 "distanceType": "A String", # Distance type for clustering models.
1521 "kmeansInitializationColumn": "A String", # The column used to provide the initial centroids for kmeans algorithm
1522 # when kmeans_initialization_method is CUSTOM.
1523 "l2Regularization": 3.14, # L2 regularization coefficient.
1524 "dropout": 3.14, # Dropout probability for dnn models.
1525 "minRelativeProgress": 3.14, # When early_stop is true, stops training when accuracy improvement is
1526 # less than 'min_relative_progress'. Used only for iterative training
1527 # algorithms.
1528 "l1Regularization": 3.14, # L1 regularization coefficient.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1529 "maxIterations": "A String", # The maximum number of iterations in training. Used only for iterative
1530 # training algorithms.
1531 "earlyStop": True or False, # Whether to stop early when the loss doesn't improve significantly
1532 # any more (compared to min_relative_progress). Used only for iterative
1533 # training algorithms.
1534 "initialLearnRate": 3.14, # Specifies the initial learning rate for the line search learn rate
1535 # strategy.
1536 "dataSplitColumn": "A String", # The column to split data with. This column won't be used as a
1537 # feature.
1538 # 1. When data_split_method is CUSTOM, the corresponding column should
1539 # be boolean. The rows with true value tag are eval data, and the false
1540 # are training data.
1541 # 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION
1542 # rows (from smallest to largest) in the corresponding column are used
1543 # as training data, and the rest are eval data. It respects the order
1544 # in Orderable data types:
1545 # https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data-type-properties
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1546 "numClusters": "A String", # Number of clusters for clustering models.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1547 "warmStart": True or False, # Whether to train a model from the last checkpoint.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1548 "hiddenUnits": [ # Hidden units for dnn models.
1549 "A String",
1550 ],
1551 "maxTreeDepth": "A String", # Maximum depth of a tree for boosted tree models.
1552 "userColumn": "A String", # User column specified for matrix factorization models.
1553 "kmeansInitializationMethod": "A String", # The method used to initialize the centroids for kmeans algorithm.
1554 "learnRateStrategy": "A String", # The strategy to determine learn rate for the current iteration.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1555 "dataSplitEvalFraction": 3.14, # The fraction of evaluation data over the whole input data. The rest
1556 # of data will be used as training data. The format should be double.
1557 # Accurate to two decimal places.
1558 # Default value is 0.2.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1559 "dataSplitMethod": "A String", # The data split type for training and evaluation, e.g. RANDOM.
1560 "subsample": 3.14, # Subsample fraction of the training data to grow tree to prevent
1561 # overfitting for boosted tree models.
1562 "labelClassWeights": { # Weights associated with each label class, for rebalancing the
1563 # training data. Only applicable for classification models.
1564 "a_key": 3.14,
1565 },
1566 "learnRate": 3.14, # Learning rate in training. Used only for iterative training algorithms.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1567 "modelUri": "A String", # [Beta] Google Cloud Storage URI from which the model was imported. Only
1568 # applicable for imported models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1569 "walsAlpha": 3.14, # Hyperparameter for matrix factoration when implicit feedback type is
1570 # specified.
1571 "minSplitLoss": 3.14, # Minimum split loss for boosted tree models.
1572 "lossType": "A String", # Type of loss function used during training run.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1573 },
1574 },
1575 ],
1576 "featureColumns": [ # Output only. Input feature columns that were used to train this model.
1577 { # A field or a column.
1578 "type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly
1579 # specified (e.g., CREATE FUNCTION statement can omit the return type;
1580 # in this case the output parameter does not have this "type" field).
1581 # Examples:
1582 # INT64: {type_kind="INT64"}
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1583 # ARRAY&lt;STRING&gt;: {type_kind="ARRAY", array_element_type="STRING"}
1584 # STRUCT&lt;x STRING, y ARRAY&lt;DATE&gt;&gt;:
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1585 # {type_kind="STRUCT",
1586 # struct_type={fields=[
1587 # {name="x", type={type_kind="STRING"}},
1588 # {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}
1589 # ]}}
1590 "structType": { # The fields of this struct, in order, if type_kind = "STRUCT".
1591 "fields": [
1592 # Object with schema name: StandardSqlField
1593 ],
1594 },
1595 "arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".
1596 "typeKind": "A String", # Required. The top level type of this field.
1597 # Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").
1598 },
1599 "name": "A String", # Optional. The name of this field. Can be absent for struct fields.
1600 },
1601 ],
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1602 "labelColumns": [ # Output only. Label columns that were used to train this model.
1603 # The output of the model will have a "predicted_" prefix to these columns.
1604 { # A field or a column.
1605 "type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly
1606 # specified (e.g., CREATE FUNCTION statement can omit the return type;
1607 # in this case the output parameter does not have this "type" field).
1608 # Examples:
1609 # INT64: {type_kind="INT64"}
1610 # ARRAY&lt;STRING&gt;: {type_kind="ARRAY", array_element_type="STRING"}
1611 # STRUCT&lt;x STRING, y ARRAY&lt;DATE&gt;&gt;:
1612 # {type_kind="STRUCT",
1613 # struct_type={fields=[
1614 # {name="x", type={type_kind="STRING"}},
1615 # {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}
1616 # ]}}
1617 "structType": { # The fields of this struct, in order, if type_kind = "STRUCT".
1618 "fields": [
1619 # Object with schema name: StandardSqlField
1620 ],
1621 },
1622 "arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".
1623 "typeKind": "A String", # Required. The top level type of this field.
1624 # Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").
1625 },
1626 "name": "A String", # Optional. The name of this field. Can be absent for struct fields.
1627 },
1628 ],
1629 "creationTime": "A String", # Output only. The time when this model was created, in millisecs since the epoch.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1630 "modelType": "A String", # Output only. Type of the model resource.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1631 "encryptionConfiguration": { # Custom encryption configuration (e.g., Cloud KMS keys). This shows the
1632 # encryption configuration of the model data while stored in BigQuery
1633 # storage. This field can be used with PatchModel to update encryption key
1634 # for an already encrypted model.
1635 "kmsKeyName": "A String", # [Optional] Describes the Cloud KMS encryption key that will be used to protect destination BigQuery table. The BigQuery Service Account associated with your project requires access to this encryption key.
1636 },
1637 "modelReference": { # Required. Unique identifier for this model.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1638 "projectId": "A String", # [Required] The ID of the project containing this model.
1639 "datasetId": "A String", # [Required] The ID of the dataset containing this model.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1640 "modelId": "A String", # [Required] The ID of the model. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1641 },
1642 "etag": "A String", # Output only. A hash of this resource.
1643 "location": "A String", # Output only. The geographic location where the model resides. This value
1644 # is inherited from the dataset.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1645 "friendlyName": "A String", # Optional. A descriptive name for this model.
1646 "expirationTime": "A String", # Optional. The time when this model expires, in milliseconds since the epoch.
1647 # If not present, the model will persist indefinitely. Expired models
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1648 # will be deleted and their storage reclaimed. The defaultTableExpirationMs
1649 # property of the encapsulating dataset can be used to set a default
1650 # expirationTime on newly created models.
Dan O'Mearadd494642020-05-01 07:42:23 -0700[diff] [blame]1651 "lastModifiedTime": "A String", # Output only. The time when this model was last modified, in millisecs since the epoch.
Bu Sun Kim715bd7f2019-06-14 16:50:42 -0700[diff] [blame]1652 }</pre>
1653</div>
1654
1655</body></html>