Direct prediction model is an extension to a model with an ability to predict for single point instead of transforming dataset.
Direct prediction model is an extension to a model with an ability to predict for single point instead of transforming dataset. Used to improve performance of the combined models.
Type of the model input
Trains a model which predicts a scalar of weighted classes or a vector which is a generalized linear combination of per-class predictions
Trains a model which predicts vector (one element per class)
Collects all classes in the dataset and trains a model for each class.
Collects all classes in the dataset and trains a model for each class. Base class for multi-class and linear combination model.
Collects types, available in the dataset and trains one model per each type.
Train models for all classes.
Train models for all classes. Result is a linear combination model. Summary blocks for combined model are merged from the nested model with extra class column.
Type of the model to train.
Estimator for training the nested models.
Column where the classes are recorded.
Which classes not to include into result.
Mapping for merging/renaming classes.
Whenever to train classes in parallel.
Whenever to cache forks before iterating
Linear combination model (given an instance coputes prediction of all the models and combines them linary with configured weights).
Reader for linear combination combination model.
Train models for all classes.
Train models for all classes. Result is a vector with prediction for each class. Summary blocks for combined model are merged from the nested model with extra class column.
Type of the model to train.
Estimator for training the nested models.
Column where the classes are recorded.
Which classes not to include into result.
Mapping for merging/renaming classes.
Whenever to train classes in parallel.
Whenever to cache forks before iterating
Linear combination model (given an instance coputes prediction of all the models and combines them linary with configured weights).
Reader for multi class combination model.
Train dedicated model for a certain type.
Train dedicated model for a certain type. Result is a selector model. Summary blocks for combined model are merged from the nested model with extra type column.
Type of the model to train.
Estimator for training the nested models.
Column where the type label stored.
Whenever to train types in parallel.
Whenever to cache forks before iterating
Selector model (given instance type applies ONE of the models)
Reader for type selecting model.