import numpy as np
from sklearn.preprocessing import LabelBinarizer
class GaussianNB(object):
"""
朴素贝叶斯分类器，适用于连续型数据。
"""
@staticmethod
def gaussfunc(x, mu, singma):
"""高斯函数
:param x: 数据集
:param mu: 均值
:param singma: 方差
:return:
"""
sqsingma = singma @ singma
numerator = -np.exp(np.sum((x - mu) ** 2, axis=1) / (2 * sqsingma))
return numerator / np.sqrt(2 * np.pi * sqsingma)
def fit(self, X, y):
"""
:param X_: shape = [n_samples, n_features]
:param y: shape = [n_samples]
:return: self
"""
self.classes, self.classes_count = np.unique(y, return_counts=True)
self.mean = np.zeros((self.classes_count.shape[0],
X.shape[1]), dtype=np.float64)
self.var = np.zeros((self.classes_count.shape[0],
X.shape[1]), dtype=np.float64)
for i, label in enumerate(self.classes):
x_i = X[y == label]
self.mean[i, :] = np.mean(x_i, axis=0)
self.var[i, :] = np.var(x_i, axis=0)
return self
def predict(self, X):
"""
:param X: shape = [n_samples, n_features]
:return: shape = [n_samples]
"""
likelihood = []
for i in range(self.classes.shape[0]):
likelihood.append(self.classes_count[i] *
GaussianNB.gaussfunc(X, self.mean[i, :],
self.var[i, :]))
likelihood = np.array(likelihood).T
return np.argmax(likelihood, axis=1)
class MultinomialNB(object):
"""
朴素贝叶斯分类器，适用于离散型数据。
"""
def __init__(self, alpha=1.0):
self.alpha = alpha
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)))
def predict(self, X):
"""
:param X: shape = [n_samples, n_features]
:return: shape = [n_samples]
"""
likelihood = X @ self.feature_log_prob.T
return np.argmax(likelihood, axis=1)
