getml.predictors

This module contains machine learning algorithms to learn and predict on the generated features.

The predictor classes defined in this module serve two purposes. First, a predictor can be used as a feature_selector in Pipeline to only select the best features generated during the automated feature learning and to get rid off any redundancies. Second, by using it as a predictor, it will be trained on the features of the supplied data set and used to predict to unknown results. Every time a new data set is passed to the predict() method of one of the models, the raw relational data is interpreted in the data model, which was provided during the construction of the model, transformed into features using the trained feature learning algorithm, and, finally, its target will be predicted using the trained predictor.

The algorithms can be grouped according to their finesse and whether you want to use them for a classification or regression problem.

simple

sophisticated

regression

LinearRegression

XGBoostRegressor

classification

LogisticRegression

XGBoostClassifier

Note:

All predictors need to be passed to Pipeline.

Classes

ScaleGBMClassifier([colsample_bylevel, ...])

Standard gradient boosting classifier that fully supports memory mapping

ScaleGBMRegressor([colsample_bylevel, ...])

Standard gradient boosting regressor that fully supports memory mapping

LinearRegression([learning_rate, reg_lambda])

Simple predictor for regression problems.

LogisticRegression([learning_rate, reg_lambda])

Simple predictor for classification problems.

XGBoostClassifier([booster, ...])

Gradient boosting classifier based on xgboost.

XGBoostRegressor([booster, ...])

Gradient boosting regressor based on xgboost.