逻辑回归

Logistic回归简单分析

　　优点：计算代价不高，易于理解和实现 
　　缺点：容易欠拟合，分类精度可能不高 
　　适用数据类型：数值型和标称型数据

　　我们都知道逻辑回归是和Sigmod函数一起的，为了实现逻辑回归分类器，我们可以在每一个特征上都乘以一个回归系数，然后将所有的结果值相加，将总和代入S函数，进而得到一个范围在0~1之间的数值。任何大于0.5的数据被分人1类，小于0.5被归为0类。 
　　而现在有了分类器的函数了，那么上面提到的最佳回归系数怎么求呢？这里就出现了基于最优化方法的最佳回归系数的确定。 
　　 
　　梯度上升法：要找到某函数的最大值，最好的方法就是沿着该函数的梯度方向探寻。梯度上升法用来求函数的最大值，梯度下降法用来求函数的最小值。 

基于机器学习实战的代码

#! encoding:utf-8
from numpy import *
import matplotlib.pyplot as plt
from _socket import error
from matplotlib.font_manager import weight_as_number
from matplotlib.pyplot import scatter
from bokeh.charts.attributes import color
#加载数据
def loadDataSet():
    
    dataMat=[]
    lableMat=[]
    fr = open(r"C:\Users\QAQ\Desktop\testSet.txt")
    for line in fr.readlines():
        lineArr = line.strip().split()
        dataMat.append([1.0, float(lineArr[0]), float(lineArr[1])])
        lableMat.append(int(lineArr[2]))
    return dataMat, lableMat
#sigmoid函数
def sigmoid(intX):
    return 1.0/(1+exp(-intX))
#梯度上升优化算法
def gradAscent(dataMatIn, classLabels):
    dataMatrix = mat(dataMatIn)
    labelMat = mat(classLabels).transpose()
    m,n = shape(dataMatrix)
    alpha = 0.001
    maxCycles = 500
    weights = ones((n,1))
#     print shape(dataMatrix)
#     print shape(weights)
    for k in range(maxCycles):
        h=sigmoid(dataMatrix*weights)
        error = (labelMat - h)
        weights = weights + alpha * dataMatrix.transpose()*error
    return weights
 
#画出数据集合Logisti回归最佳拟合直线                
def plotBestFit(weights):
   
    dataMat, labelMat = loadDataSet()
    dataArr = array(dataMat)
    n = shape(dataArr)[0]
    xrecord1=[];yrecord1=[]
    xrecord2=[];yrecord2=[]
    
    for i in range(n):
        if int(labelMat[i]) == 1:
            xrecord1.append(dataArr[i,1]);yrecord1.append(dataArr[i,2])
        else:
            xrecord2.append(dataArr[i,1]);yrecord2.append(dataArr[i,2])
    
#     print dataMat
#     fig = plt.figure()
    scatter(xrecord1,yrecord1,color="red",marker='s',s=30)
    scatter(xrecord2, yrecord2,color="blue")
    x = arange(-3, 3, 1)
   
    y = (-float(weights[0])-float(weights[1])*x)/float(weights[2])
    
#     print x
#     print y
    plt.plot(x,y)
    plt.show()
    
#     dataArr = array(dataMat)
#     print dataArr
#     print type(dataArr)
             
def main():
    
  
    
    dataArr, labelMat=loadDataSet()
    plotBestFit(gradAscent(dataArr, labelMat))
#     print gradAscent(dataArr, labelMat)
main()