naive_bayes.py 文件源码

def fit(self, X, y):
        """
        :param X_: shape = [n_samples, n_features] 
        :param y: shape = [n_samples] 
        :return: self
        """
        labelbin = LabelBinarizer()
        Y = labelbin.fit_transform(y)
        self.classes = labelbin.classes_
        self.class_count = np.zeros(Y.shape[1], dtype=np.float64)
        self.feature_count = np.zeros((Y.shape[1], X.shape[1]),
                                      dtype=np.float64)

        self.feature_count += Y.T @ X
        self.class_count += Y.sum(axis=0)
        smoothed_fc = self.feature_count + self.alpha
        smoothed_cc = smoothed_fc.sum(axis=1)

        self.feature_log_prob = (np.log(smoothed_fc) -
                                 np.log(smoothed_cc.reshape(-1, 1)))