ensembles.py 文件源码

def fit(self, features, targets):
        """
        Trains self.nb_trees number of decision trees.
        :param features:    Array-like object of shape (nb_samples, nb_features)
                            containing the training examples
        :param targets:     Array-like object of shape (nb_samples) containing the
                            training targets.
        """
        if not self.nb_samples:
            self.nb_samples = int(len(features) / 10)
        with ProcessPoolExecutor(max_workers=self.max_workers) as executor:
            random_features = []
            for x in range(self.nb_trees):
                idxs = np.random.choice(np.arange(len(features)), self.nb_samples, replace=True)
                try:
                    chosen_features = itemgetter(*idxs)(features)
                    chosen_targets = itemgetter(*idxs)(targets)
                except:
                    chosen_features = features.iloc[idxs].as_matrix()
                    chosen_targets = targets.iloc[idxs].as_matrix()
                random_features.append((x, chosen_features, chosen_targets))
            self.trees = list(executor.map(self.train_tree, random_features))