API Reference

Estimators

Estimators are one of the stages that are added to a pipeline. An estimator is typically the last stage in a pipeline.

When you add an estimator, such as a DecisionTree or LinearRegression to a pipeline, the fit method performs parameter estimation and generates a model. The pipeline returns a PipelineModel that includes the model.

Decision Tree

DecisionTree([alpha, cf_level, criterion, ...]) Decision Tree
DecisionTree.fit(table, *args, **kwargs) Fit function for decision tree

Decision Forest

DecisionForest([alpha, bootstrap, cf_level, ...]) Decision Forest
DecisionForest.fit(table, *args, **kwargs) Fit function for decision forest

Gradient Boosting Tree

GBTree([distribution, ...]) Gradient Boosting Tree
GBTree.fit(table, *args, **kwargs) Fit function for gradient boosting tree

Logistic Regression

LogisticRegression([intercept, max_effects, ...]) Logistic Regression
LogisticRegression.fit(table, *args, **kwargs) Fit function for logistic regression

Linear Regression

LinearRegression([intercept, max_effects, ...]) Linear Regression
LinearRegression.fit(table, *args, **kwargs) Fit function for linear regression

Neural Network

NeuralNetwork([acts, annealing_rate, ...]) Neural Network
NeuralNetwork.fit(table, *args, **kwargs) Fit function for neural network

Models

The model classes are created as a result of adding an estimator class to a Pipeline and running the fit method with training data. The pipeline returns a PipelineModel that includes the model form of the estimator. For example, if the pipeline included a NeuralNetwork estimator, then the returned pipeline model includes a NeuralNetworkModel instance.

Decision Tree Model

DecisionTreeModel.score(table, *args, **kwargs) Score the data using the model
DecisionTreeModel.transform(table, *args, ...) Transform function for the model

Decision Forest Model

DecisionForestModel.score(table, *args, **kwargs) Score the data using the model
DecisionForestModel.transform(table, *args, ...) Transform function for the model

Gradient Boosting Tree Model

GBTreeModel.score(table, *args, **kwargs) Score the data using the model
GBTreeModel.transform(table, *args, **kwargs) Transform function for the model

Logistic Regression Model

LogisticRegressionModel.score(table, *args, ...) Score the data using the model
LogisticRegressionModel.transform(table, ...) Transform function for the model

Linear Regression Model

LinearRegressionModel.score(table, *args, ...) Score the data using the model
LinearRegressionModel.transform(table, ...) Transform function for the model

Neural Network Model

NeuralNetworkModel.score(table, *args, **kwargs) Score the data using the model
NeuralNetworkModel.transform(table, *args, ...) Transform function for the model

Pipelines

Pipeline

Pipelines are a series of transformers followed by an estimator used to construct a self-contained workflow. When the fit or score method of the pipeline is executed, each stage of the pipeline is executed in order. The output of each stage is used as the input for the next stage. The result is the output from the last stage.

Pipeline(stages) Execute a series of transformers and estimators
Pipeline.fit(table, *args, **kwargs) Train the models using the stages in the pipeline
Pipeline.transform(table, *args, **kwargs) Execute the transformations in this pipeline only

Pipeline Model

PipelineModel.score(table, **kwargs) Apply transformations and score the data using the trained model
PipelineModel.transform(table) Run the transforms in the trained pipeline

Transformers

Transformers are used to modify your data sets. Currently, this includes various forms of imputing missing values in your data sets.

Imputers

Imputer([value]) Impute missing values in a data set
Imputer.transform(table[, value]) Perform the imputation on the given data set

HyperParameter Tuning

Hyperparameter tuning allows you to test various combinations of model parameters in one workflow. This can be done either on a single Estimator class instance or a Pipeline.

HyperParameterTuning(**kwargs) Perform search over all combinations of specified parameters
HyperParameterTuning.gridsearch(table[, n_jobs]) Fit model over various permutations of parameters