Logistic回归梯度上升分类法

#coding=utf-8
from numpy import *
import re
def load_data():
    """加载数据"""
    data=[];label=[]
    #附加两个表
    open_file=open('test.txt')
    for line in open_file.readlines():
        line_arr=re.split(r'(\d*)',line.strip())
        data.append([1.0,float(int(line_arr[1])),float(int(line_arr[3]))])
        label.append(int(line_arr[5]))
    #这个for就是附加表用的，里面的正则是因为数据格式问题。也就是对数据进行筛选
    return data,label
def sigmoid(inx):
    """sigmoid函数"""
    return 1.0/(1+exp(-inx))
def grad_ascent(data,label):
    """回归梯度上升"""
    data_matrix=mat(data)
    #把原始数据矩阵化
    label_matrix=mat(label).T
    #矩阵化后转置
    m,n=shape(data_matrix)
    #m代表行，n代表列
    alpha=0.001
    #步幅
    max_cycles=5000
    #迭代次数,本列中迭代次数在阈值内次数越多越精确
    weights=ones((n,1))
    #weights[0]步长，weights[1]&weights[2]决定了拟合直线的斜率
    for k in range(max_cycles):
        h=sigmoid(data_matrix*weights)
        #data_matrix*weights就相当于求data_matrix中每行元素与weights每列元素乘积的和
        error=(label_matrix-h)
        #类别目标的差值。即元素a经sigmoid判定后为类别1.它的sigmoid值为0.72 。error=1-0.72        
        weights=weights+alpha*data_matrix.transpose()*error
        #w=w+α▽wf（w）梯度上升算法迭代公式。按定义看
        #梯度上升，weights[0]为累计步长        
    weights=weights.getA()
    return weights
def best_fit(weights):
    """"""
    import matplotlib.pyplot as plt
    data,label=load_data()
    data_arr=array(data)
    #阵列化
    m=shape(data_arr)[0]
    #获取数据行数
    xcord1=[];ycord1=[]
    xcord0=[];ycord0=[]
    #初始列表
    for i in range(m):
        if int(label[i])==1:
            xcord1.append(data_arr[i,1])
            ycord1.append(data_arr[i,2])
            #获取1类坐标
        else:
            xcord0.append(data_arr[i,1])
            ycord0.append(data_arr[i,2])
            #获取0类坐标
    fig=plt.figure()
    #创建图形
    ax=fig.add_subplot(111)
    #创建一个子图
    ax.scatter(xcord1,ycord1,s=30,c='red',marker='s')
    ax.scatter(xcord0,ycord0,s=30,c='green')
    #创建散点图
    x=arange(8.0,10.0,0.1)
    #拟合直线x轴。前两个参数是范围，后一个是精度。
    #关于范围的选定应该看两类数据x轴的重合范围。
    #如样本数据是只在9处重合，那取(8，10)这个区间。
    y=(-weights[0]-weights[1]*x)/weights[2]
    ax.plot(x,y)
    #创建最佳拟合直线
    plt.xlabel('X1')
    plt.ylabel('X2')
    plt.show()
    

样本数据：

[2, 4, '1']
[5, 8, '1']
[8, 5, '1']
[9, 9, '1']
[4, 1, '1']
[0, 0, '1']
[5, 8, '1']
[9, 3, '1']
[1, 8, '1']
[7, 3, '1']
[9, 3, '1']
[3, 8, '1']
[4, 6, '1']
[9, 7, '1']
[7, 1, '1']
[5, 2, '1']
[9, 6, '1']
[6, 9, '1']
[9, 8, '1']
[7, 0, '1']
[4, 5, '1']
[9, 8, '1']
[0, 4, '1']
[4, 3, '1']
[6, 0, '1']
[9, 9, '1']
[0, 3, '1']
[9, 8, '1']
[1, 7, '1']
[5, 8, '1']
[7, 8, '1']
[1, 5, '1']
[0, 7, '1']
[1, 9, '1']
[7, 8, '1']
[2, 5, '1']
[7, 4, '1']
[2, 1, '1']
[6, 1, '1']
[0, 1, '1']
[2, 4, '1']
[6, 0, '1']
[8, 0, '1']
[4, 9, '1']
[8, 3, '1']
[9, 8, '1']
[8, 9, '1']
[5, 9, '1']
[9, 6, '1']
[4, 2, '1']
[8, 7, '1']
[1, 9, '1']
[3, 8, '1']
[0, 1, '1']
[1, 1, '1']
[0, 9, '1']
[0, 6, '1']
[1, 5, '1']
[2, 6, '1']
[9, 5, '1']
[5, 0, '1']
[2, 4, '1']
[5, 9, '1']
[9, 5, '1']
[6, 3, '1']
[9, 3, '1']
[3, 6, '1']
[8, 6, '1']
[7, 7, '1']
[0, 0, '1']
[5, 4, '1']
[2, 9, '1']
[5, 7, '1']
[3, 9, '1']
[6, 9, '1']
[8, 2, '1']
[8, 3, '1']
[8, 0, '1']
[2, 4, '1']
[9, 2, '1']
[0, 3, '1']
[6, 8, '1']
[5, 4, '1']
[5, 0, '1']
[5, 3, '1']
[7, 6, '1']
[0, 4, '1']
[3, 9, '1']
[7, 5, '1']
[8, 3, '1']
[9, 7, '1']
[8, 3, '1']
[3, 5, '1']
[2, 6, '1']
[1, 9, '1']
[6, 2, '1']
[3, 5, '1']
[9, 7, '1']
[5, 6, '1']
[7, 2, '1']
[11, 0, '0']
[13, 7, '0']
[16, 5, '0']
[11, 0, '0']
[17, 6, '0']
[9, 5, '0']
[15, 1, '0']
[13, 7, '0']
[12, 6, '0']
[9, 5, '0']
[17, 4, '0']
[10, 8, '0']
[10, 8, '0']
[9, 5, '0']
[13, 2, '0']
[13, 6, '0']
[9, 0, '0']
[11, 9, '0']
[17, 2, '0']
[9, 7, '0']
[16, 4, '0']
[12, 1, '0']
[10, 8, '0']
[10, 1, '0']
[17, 7, '0']
[12, 0, '0']
[16, 5, '0']
[18, 2, '0']
[15, 6, '0']
[9, 5, '0']
[13, 4, '0']
[13, 2, '0']
[10, 2, '0']
[17, 7, '0']
[16, 1, '0']
[15, 0, '0']
[9, 4, '0']
[16, 7, '0']
[13, 1, '0']
[17, 0, '0']
[18, 4, '0']
[12, 3, '0']
[10, 7, '0']
[14, 6, '0']
[9, 5, '0']
[11, 9, '0']
[12, 4, '0']
[17, 8, '0']
[10, 2, '0']
[12, 5, '0']
[13, 0, '0']
[12, 2, '0']
[11, 1, '0']
[14, 1, '0']
[17, 0, '0']
[18, 3, '0']
[10, 5, '0']
[18, 2, '0']
[12, 4, '0']
[15, 8, '0']
[17, 9, '0']
[18, 5, '0']
[14, 9, '0']
[16, 9, '0']
[18, 5, '0']
[9, 1, '0']
[14, 4, '0']
[13, 2, '0']
[12, 9, '0']
[16, 8, '0']
[15, 4, '0']
[12, 0, '0']
[16, 9, '0']
[14, 3, '0']
[12, 9, '0']
[17, 5, '0']
[11, 4, '0']
[13, 6, '0']
[16, 3, '0']
[16, 2, '0']
[11, 3, '0']
[11, 1, '0']
[17, 9, '0']
[18, 2, '0']
[11, 8, '0']
[14, 3, '0']
[11, 0, '0']
[18, 6, '0']
[12, 6, '0']
[10, 0, '0']
[14, 0, '0']
[16, 5, '0']
[12, 7, '0']
[15, 0, '0']
[15, 1, '0']
[18, 9, '0']
[9, 0, '0']
[18, 0, '0']
[18, 6, '0']
[9, 3, '0']