image.png
    image.png
    image.png

    1. def loadDataSet():
    2.     dataMat = []; labelMat = [] 
    3.     fr = open('4-Logistic/testSet.txt') 
    4.     for line in fr.readlines():
    5.         lineArr = line.strip().split()
    6.         dataMat.append([1.0, float(lineArr[0]), float(lineArr[1])])
    7.         labelMat.append(int(lineArr[2])) 
    8.     fr.close() 
    9.     return dataMat, labelMat
    11. def sigmoid(inX):
    12.     return 1.0 / (1 + np.exp(-inX))
    14. def gradAscent(dataMatIn, classLabels):
    15.     dataMatrix = np.mat(dataMatIn) 
    16.     labelMat = np.mat(classLabels).transpose()
    17.     m, n = np.shape(dataMatrix) 
    18.     alpha = 0.001
    19.     maxCycles = 500
    20.     weights = np.ones((n,1))
    21.     for k in range(maxCycles):
    22.         h = sigmoid(dataMatrix * weights)
    23.         error = labelMat - h
    24.         weights = weights + alpha * dataMatrix.transpose() * error
    25.     return weights.getA()
    26. def plotBestFit(wei):
    27.     dataMat, labelMat = loadDataSet() 
    28.     dataArr = np.array(dataMat)
    29.     n = np.shape(dataMat)[0]
    30.     xcord1 = []; ycord1 = []
    31.     xcord2 = []; ycord2 = []
    32.     for i in range(n):
    33.     if int(labelMat[i]) == 1:
    34.             xcord1.append(dataArr[i,1]); ycord1.append(dataArr[i,2])
    35.     else:
    36.             xcord2.append(dataArr[i,1]); ycord2.append(dataArr[i,2])
    37.     fig = plt.figure()
    38.     ax = fig.add_subplot(111) 
    39.     ax.scatter(xcord1, ycord1, s = 20, c = 'red', marker = 's',alpha=.5)
    40.     ax.scatter(xcord2, ycord2, s = 20, c = 'green',alpha=.5)            
    41.     x = np.arange(-3.0, 3.0, 0.1)
    42.     y = (-wei[0] - wei[1] * x) / wei[2]
    43.     ax.plot(x, y)
    44.     plt.title('BestFit') 
    45.     plt.xlabel('X1'); plt.ylabel('X2')
    46.     plt.show()
    48. def stocGradAscentO(dataMatrix, classLabels):     
    49.     dataMatrix=np.array(dataMatrix)
    50.     m,n = np.shape(dataMatrix)
    51.     alpha = 0.01
    52.     weights = np.ones(n)
    53.     for i in range(m):
    54.         h = sigmoid(sum(dataMatrix[i]*weights))
    55.         error = classLabels[i] - h
    56.         weights = weights + alpha * error * dataMatrix[i]
    57.     return weights
    59. def stocGradAscent1(dataMatrix, classLabels, numIter=150):
    60.     dataMatrix=np.array(dataMatrix)
    61.     m,n = np.shape(dataMatrix) 
    62.     weights = np.ones(n)
    63.     for j in range(numIter):                                           
    64.         dataIndex = list(range(m))
    65.         for i in range(m): 
    66.             alpha = 4/(1.0+j+i)+0.01 
    67.             randIndex = int(random.uniform(0,len(dataIndex)))
    68.             h = sigmoid(sum(dataMatrix[randIndex]*weights))
    69.             error = classLabels[randIndex] - h 
    70.             weights = weights + alpha * error * dataMatrix[randIndex]
    71.             del(dataIndex[randIndex])
    72.     return weights