arbitragelab.ml_approach.regressor_committee

Regressor Committee.

Module Contents

Classes

RegressorCommittee

Regressor Committee implementation which basically fits N number of models
class RegressorCommittee(regressor_params: dict, regressor_class: str = 'MultiLayerPerceptron', num_committee: int = 10, epochs: int = 20, patience: int = 100, verbose: bool = True)

Regressor Committee implementation which basically fits N number of models and takes the mean value of their predictions.

fit(xtrain: pandas.DataFrame, ytrain: pandas.DataFrame, xtest: pandas.DataFrame, ytest: pandas.DataFrame)

Fits the member models, then the voting object.

Parameters:

  • xtrain – (pd.DataFrame) Input training data.

  • ytrain – (pd.DataFrame) Target training data.

  • xtest – (pd.DataFrame) Input test data.

  • ytest – (pd.DataFrame) Target test data.

predict(xtest: pandas.DataFrame) → pandas.DataFrame

Collects results from all the committee members and returns average result.

Parameters:

xtest – (pd.DataFrame) Input test data.

Returns:

(pd.DataFrame) Model predictions.

plot_losses(figsize: tuple = (15, 8)) → matplotlib.axes._axes.Axes

Plot all individual member loss metrics.

Parameters:

figsize – (tuple)

Returns:

(Axes)