logistic_regression.py

#!/usr/bin/env python
# encoding: utf-8

from numpy import *
import matplotlib.pyplot as plt

def logistic(wTx):
    return 1.0/(1.0 + exp(-wTx))

def file2matrix(filename,delimiter):
	recordlist = []
	fp = open(filename,"rb")
	content = fp.read()
	fp.close()
	rowlist = content.splitlines()
	# 逐行遍历 	并将结果按分隔符分割为行向量
	recordlist=[map(eval, row.split(delimiter)) for row in rowlist if row.strip()]
    # 返回转换后的矩阵形式
	return mat(recordlist)

def buildMat(dataSet):
	m,n=shape(dataSet)
	dataMat = zeros((m,n))
	dataMat[:,0] = 1
	dataMat[:,1:] = dataSet[:,:-1]
	return 	dataMat

def load_data():
    filename = "testSet.txt"
    Input = file2matrix(filename, delimiter='\t')
    label = Input[:,-1]
    return Input, label

def logistic_regression_bgd(Input, label):
    num_samples, num_features = shape(Input)
    dataMat = buildMat(Input)
    alpha = 0.001
    steps = 500
    weights = ones((num_features, 1))
    for iteration in range(steps):
        output = logistic(dataMat * mat(weights))
        error = label - output
        weights = weights + alpha*dataMat.T*error
    print weights
    return weights


def logistic_regression_sgd(Input, label):
    num_samples, num_features = shape(Input)
    dataMat = buildMat(Input)
    alpha = 0.001
    steps = 500
    weights = ones(num_features)
    for iteration in range(steps):
        dataIndex = range(num_samples)
        for i in range(num_samples):
            randIndex = random.randint(0, len(dataIndex))
            output = logistic(sum(dataMat[randIndex]*weights.T))
            error = label[randIndex] - output
            weights = weights + alpha*error*dataMat[randIndex]
            del(dataIndex[randIndex])
    print weights
    return weights

def logistic_regression_nsgd(Input, label):
    num_samples, num_features = shape(Input)
    dataMat = buildMat(Input)
    # alpha = 0.001
    steps = 500
    weights = ones(num_features)
    for iteration in range(steps):
        dataIndex = range(num_samples)
        for i in range(num_samples):
            alpha = 1.0/(1.0+i+iteration) + 0.0001
            randIndex = random.randint(0, len(dataIndex))
            output = logistic(sum(dataMat[randIndex]*weights.T))
            error = label[randIndex] - output
            weights = weights + alpha*error*dataMat[randIndex]
            del(dataIndex[randIndex])
    print weights
    return weights


if __name__ == '__main__':
    Input, label = load_data()
    weights = logistic_regression_bgd(Input, label)
    weights_2 = logistic_regression_sgd(Input, label)
    weights_3 = logistic_regression_nsgd(Input, label)