LogisticRegression

python 实现logisticRegression

1、实验使用的数据来自http://sci2s.ugr.es/keel/category.php?cat=clas
2、此处并没有考虑正则项，如果需要考虑正则，只需要修改梯度的计算
3、有关具体的推导请参考Andrew Ng.的课程http://cs229.stanford.edu/

#-*- coding=utf-8 -*-

import numpy as np
"""
time 2016.1.31
"""
class LogisticRegression(object):
    "class logistic Regression"
    def __init__(self,n,epo=1000,rate=0.001):
        self.rate=rate
        self.epo=epo
        self.weights=np.random.normal(size=(n))
    def fit(self,data):
        for i in range(self.epo):
            for j in range(data.shape[0]):
                x=data[j][:-1]
                y=data[j][-1]
                h=self.predict(x)
                self.weights=self.weights+self.rate*(y-h)*x
    def sigmoid(self,x):
        return 1/(1+np.exp(-x))
    def predict(self,x):
        g=sum(self.weights*x)
        return self.sigmoid(g)
    def accuracy(self,data):
        num=0
        for i in range(data.shape[0]):
            x=data[i][:-1]
            y=data[i][-1]
            h=self.predict(x)
            if h>=0.5:
                if(y==1):
                    num+=1
            if h<0.5:
                if y==0:
                    num+=1
        print "predict accuracy is %lf"%(num*1.0/data.shape[0])

def loadData(path):
    data=np.loadtxt(path,skiprows=10,dtype="int32",delimiter=",")
    new_data=np.ones((data.shape[0],data.shape[1]+1))
    """"relarge the dataset with x0=1"""
    new_data[:,1:]=data
    return new_data
data=loadData("D:\\SelfLearning\\Machine Learning\\ClassifyDataSet\\mammographic\\mammographic.dat")
train_data=data[:600]
test_data=data[600:]
lr=LogisticRegression(6,epo=1000)
lr.fit(train_data)

print lr.weights
lr.accuracy(test_data)
lr.accuracy(train_data)