python-实现朴素贝叶斯分类器

本文用的是sciki-learn库的iris数据集进行测试。用的模型也是最简单的，就是用贝叶斯定理
P(A|B) = P(B|A)*P(A)/P(B)，计算每个类别在样本中概率（代码中是pLabel变量）
以及每个类下每个特征的概率（代码中是pNum变量）。写得比较粗糙，对于某个类下没有此特征的情况采用p=1/样本数量。

# -*- coding:utf-8 -*- 
from numpy import * 
from sklearn import datasets 
import numpy as np 

class NaiveBayesClassifier(object): 

  def __init__(self): 
    self.dataMat = list() 
    self.labelMat = list() 
    self.pLabel = {} 
    self.pNum = {} 

  def loadDataSet(self): 
    iris = datasets.load_iris() 
    self.dataMat = iris.data 
    self.labelMat = iris.target 
    labelSet = set(iris.target) 
    labelList = [i for i in labelSet] 
    labelNum = len(labelList) 
    for i in range(labelNum): 
      self.pLabel.setdefault(labelList[i]) 
      self.pLabel[labelList[i]] = np.sum(self.labelMat==labelList[i])/float(len(self.labelMat)) 

  def seperateByClass(self): 
    seperated = {} 
    for i in range(len(self.dataMat)): 
      vector = self.dataMat[i] 
      if self.labelMat[i] not in seperated: 
        seperated[self.labelMat[i]] = [] 
      seperated[self.labelMat[i]].append(vector) 
    return seperated 

  # 通过numpy array二维数组来获取每一维每种数的概率 
  def getProbByArray(self, data): 
    prob = {} 
    for i in range(len(data[0])): 
      if i not in prob: 
        prob[i] = {} 
      dataSetList = list(set(data[:, i])) 
      for j in dataSetList: 
        if j not in prob[i]: 
          prob[i][j] = 0 
        prob[i][j] = np.sum(data[:, i] == j) / float(len(data[:, i])) 
    prob[0] = [1 / float(len(data[:,0]))] # 防止feature不存在的情况 
    return prob 

  def train(self): 
    featureNum = len(self.dataMat[0]) 
    seperated = self.seperateByClass() 
    t_pNum = {} # 存储每个类别下每个特征每种情况出现的概率 
    for label, data in seperated.iteritems(): 
      if label not in t_pNum: 
        t_pNum[label] = {} 
      t_pNum[label] = self.getProbByArray(np.array(data)) 
    self.pNum = t_pNum 

  def classify(self, data): 
    label = 0 
    pTest = np.ones(3) 
    for i in self.pLabel: 
      for j in self.pNum[i]: 
        if data[j] not in self.pNum[i][j]: 
          pTest[i] *= self.pNum[i][0][0] 
        else: 
          pTest[i] *= self.pNum[i][j][data[j]] 
    pMax = np.max(pTest) 
    ind = np.where(pTest == pMax) 
    return ind[0][0] 

  def test(self): 
    self.loadDataSet() 
    self.train() 
    pred = [] 
    right = 0 
    for d in self.dataMat: 
      pred.append(self.classify(d)) 
    for i in range(len(self.labelMat)): 
      if pred[i] == self.labelMat[i]: 
        right += 1 
    print right / float(len(self.labelMat)) 

if __name__ == '__main__': 
  NB = NaiveBayesClassifier() 
  NB.test() 

#输出结果：
0.96

https://www.oudahe.com/