Train and test a cross-validated supervised learning regressor.
Predicts a continuous sample metadata column using a supervised learning
regressor. Splits input data into training and test sets. The training set
is used to train and test the estimator using a stratified k-fold cross-
validation scheme. This includes optional steps for automated feature
extraction and hyperparameter optimization. The test set validates
classification accuracy of the optimized estimator. Outputs classification
results for test set. For more details on the learning algorithm, see
http://scikit-learn.org/stable/supervised_learning.html
Parameters
- table : FeatureTable[Frequency]
- Feature table containing all features that should be used for target
prediction.
- metadata : MetadataColumn[Numeric]
- Numeric metadata column to use as prediction target.
- test_size : Float % Range(0.0, 1.0, inclusive_start=False), optional
- Fraction of input samples to exclude from training set and use for
classifier testing.
- step : Float % Range(0.0, 1.0, inclusive_start=False), optional
- If optimize_feature_selection is True, step is the percentage of
features to remove at each iteration.
- cv : Int % Range(1, None), optional
- Number of k-fold cross-validations to perform.
- random_state : Int, optional
- Seed used by random number generator.
- n_estimators : Int % Range(1, None), optional
- Number of trees to grow for estimation. More trees will improve
predictive accuracy up to a threshold level, but will also increase
time and memory requirements. This parameter only affects ensemble
estimators, such as Random Forest, AdaBoost, ExtraTrees, and
GradientBoosting.
- estimator : Str % Choices('RandomForestRegressor', 'ExtraTreesRegressor', 'GradientBoostingRegressor', 'AdaBoostRegressor', 'ElasticNet', 'Ridge', 'Lasso', 'KNeighborsRegressor', 'LinearSVR', 'SVR'), optional
- Estimator method to use for sample prediction.
- optimize_feature_selection : Bool, optional
- Automatically optimize input feature selection using recursive feature
elimination.
- stratify : Bool, optional
- Evenly stratify training and test data among metadata categories. If
True, all values in column must match at least two samples.
- parameter_tuning : Bool, optional
- Automatically tune hyperparameters using random grid search.
- missing_samples : Str % Choices('error', 'ignore'), optional
- How to handle missing samples in metadata. "error" will fail if missing
samples are detected. "ignore" will cause the feature table and
metadata to be filtered, so that only samples found in both files are
retained.
Returns
- sample_estimator : SampleEstimator[Regressor]
- Trained sample estimator.
- feature_importance : FeatureData[Importance]
- Importance of each input feature to model accuracy.
- predictions : SampleData[RegressorPredictions]
- Predicted target values for each input sample.
- model_summary : Visualization
- Summarized parameter and (if enabled) feature selection information for
the trained estimator.
- accuracy_results : Visualization
- Accuracy results visualization.