Source code for traveltimes_prediction.models.algorithms.elastic_net_wrapper

from sklearn.linear_model import ElasticNet
from sklearn.preprocessing import StandardScaler
import numpy as np

from ..base_model import BaseModel


[docs]class ElasticNetWrapper(ElasticNet, BaseModel): """ Class - wrapper for ElasticNet. """ name='ElasticNet' def __init__(self, **kwargs): """ Constructor. :param dict kwargs: """ super().__init__(**kwargs) self.scaler = StandardScaler()
[docs] def fit(self, X, y, sample_weigth=None): """ Method for fitting of the estimator. :param numpy.ndarray X: Matrix of input features. :param numpy.ndarray y: Vector of ground truths for X. :param numpy.ndarray sample_weigth: :return: object - self """ self._get_descriptors(X) self.scaler = self.scaler.fit(X) X = self.scaler.transform(X) self.min_y_train = np.percentile(y, 10) self.max_y_train = max(y) # ElasticNet requires Fortran style arrays X = np.asfortranarray(X) y = np.asfortranarray(y) super().fit(X, y) return self
[docs] def predict(self, X): """ Method for prediction. :param numpy.ndarray X: Data, which should be used as input for the estimation of predicted value. :return: np.ndarray """ X = self._impute_prediction_sample(X=X) X = self.scaler.transform(X) X = np.asfortranarray(X) pred = super().predict(X=X) return self._coerce(pred)
[docs] def dump(self): """ Method for dumping of the model providing descriptors of model in dict. :return: dict """ d = dict() d['model'] = dict() d['model']['coef_'] = self.coef_.tolist() d['model']['intercept_'] = self.intercept_.tolist() d['model']['n_iter_'] = self.n_iter_ d['model']['min_y_train'] = self.min_y_train d['model']['max_y_train'] = self.max_y_train d['model']['scaler'] = dict() d['model']['scaler']['scale_'] = self.scaler.scale_.tolist() d['model']['scaler']['mean_'] = self.scaler.mean_.tolist() d['model']['scaler']['var_'] = self.scaler.var_.tolist() d['model']['scaler']['n_samples_seen_'] = self.scaler.n_samples_seen_ d['model']['_median_imputer'] = self._median_imputer.tolist() d['model_type'] = self.name return d
@staticmethod
[docs] def load(model): """ Method for loading of the model - recreating it from dumped dict. :param dict model: dumped model :return: ElasticNetWrapper """ inst = ElasticNetWrapper() inst.coef_ = np.array(model['coef_']) inst.intercept_ = np.array(model['intercept_']) inst.n_iter_ = model['n_iter_'] inst.min_y_train = model['min_y_train'] inst.max_y_train = model['max_y_train'] inst.scaler.scale_ = np.array(model['scaler']['scale_']) inst.scaler.mean_ = np.array(model['scaler']['mean_']) inst.scaler.var_ = np.array(model['scaler']['var_']) inst.scaler.n_samples_seen_d = model['scaler']['n_samples_seen_'] inst._median_imputer = np.array(model['_median_imputer']) return inst