logistic

数据的可视化和梯度上升法

import matplotlib.pyplot as plt


定义数据集下载函数

def loadDataSet():dataMat &＃61; []labelMat &＃61; []f &＃61; open(&＃39;testSet.txt&＃39;)for line in f.readlines():line_list &＃61; line.strip().split()dataMat.append([1,float(line_list[0]),float(line_list[1])])labelMat.append(int(line_list[2]))return dataMat,labelMat


定义激活函数

def sigmoid(inX):return 1.0 / (1&＃43;np.exp(-inX))


import numpy as np
dataMat,labelMat &＃61; loadDataSet()
dataMatArr &＃61; np.array(dataMat)
n &＃61; np.shape(dataArr)[0]
x_cord_1 &＃61; []; y_cord_1 &＃61; []
x_cord_2 &＃61; []; y_cord_2 &＃61; []
for i in range(n):if int(labelMat[i]) &＃61;&＃61; 1:x_cord_1.append(dataMatArr[i,1])y_cord_1.append(dataMatArr[i,2])else:x_cord_2.append(dataMatArr[i,1])y_cord_2.append(dataMatArr[i,2])


绘制图像

fig &＃61; plt.figure()
ax &＃61; fig.add_subplot(111) # 将画布分成一行一列&＃xff0c;按从左到右&＃xff0c;从上到下去一个格子
ax.scatter(x_cord_1,y_cord_1,s&＃61;30,c&＃61;&＃39;red&＃39;)
ax.scatter(x_cord_2,y_cord_2,s&＃61;30,c&＃61;&＃39;green&＃39;)


实现梯度上升算法函数

def gradAscent(dataMatIn, classLabels):# 获取数据&＃xff0c;并转化成dataMatrix &＃61; np.mat(dataMatIn)labelMatrix &＃61; np.mat(classLabels).transpose()m,n &＃61; np.shape(dataMatrix)alpha &＃61; 0.001maxCycles &＃61; 500weights &＃61; np.ones((n,1))for k in range(maxCycles):h &＃61; sigmoid(dataMatrix * weights)error &＃61; (labelMatrix - h)weights &＃61; weights &＃43; alpha * dataMatrix.transpose() * errorreturn weights


weights &＃61; gradAscent(dataArr,labelMat)
weights


matrix([[ 4.12414349],[ 0.48007329],[-0.6168482 ]])


分析数据&＃xff0c;画出决策边界

def plotBestFit(weights):dataMat,labelMat &＃61; loadDataSet()dataArr &＃61; np.array(dataMat)n &＃61; np.shape(dataArr)[0]x_cord_1 &＃61; []; y_cord_1 &＃61; []x_cord_2 &＃61; []; y_cord_2 &＃61; []for i in range(n):if int(labelMat[i]) &＃61;&＃61; 1:x_cord_1.append(dataArr[i,1])y_cord_1.append(dataArr[i,2])else:x_cord_2.append(dataArr[i,1])y_cord_2.append(dataArr[i,2])fig &＃61; plt.figure()ax &＃61; fig.add_subplot(111)ax.scatter(x_cord_1,y_cord_1,s&＃61;30,c&＃61;&＃39;red&＃39;)ax.scatter(x_cord_2,y_cord_2,s&＃61;30,c&＃61;&＃39;green&＃39;)x &＃61; np.arange(-3,3,0.1)y &＃61; (-weights[0]-weights[1]*x) / weights[2] # w0 * x0 &＃43; w1 * x1 &＃43; w2 * x2 &＃61; 0ax.plot(x,y)plt.xlabel(&＃39;X1&＃39;)plt.ylabel(&＃39;X2&＃39;)plt.show()


plotBestFit(weights.getA()) # getA()函数与mat()函数的功能相反&＃xff0c;是将一个numpy矩阵转换为数组