逻辑回归----梯度上升

逻辑回归----梯度上升
测试代码
import math
def sigmoid(data, weight):
    z = sum([data[i]*weight[i] for i in range(len(data))])
    try:
        return 1.0/(1+math.exp(-z))
    except:
        if z > 0: return 1.0
        else: return 0.0

def logistic_classify(data, weight):
    prob = sigmoid(data, weight)
    if prob > 0.5: return 1.0
    else: return 0.0

def grad_ascent(dataset, datalabel):
    weight = [1 for i in range(len(dataset[0]))]
    alpha = 0.01
    for k in range(500):
        errset = []
        for i in range(len(dataset)):
            sig = sigmoid(dataset[i], weight)
            errset.append(datalabel[i]-sig)

        for i in range(len(dataset[0])):
            for j in range(len(dataset)):
                weight[i] += alpha*dataset[j][i]*errset[j]
    return weight

def rand_grad_ascent(dataset, datalabel):
    weight = [1 for i in range(len(dataset[0]))]
    alpha = 0.01
    for i in range(len(dataset)):
        sig = sigmoid(dataset[i], weight)
        err = datalabel[i] - sig
        for j in range(len(weight)):
            weight[j] += alpha*err*dataset[i][j]

    return weight

def test(class_func):
    f_train = open('horseColicTraining.txt')
    f_test = open('horseColicTest.txt')

    trainset, trainlabel = [], []
    for line in f_train.readlines():
        line_cur = line.strip().split('	')
        trainset.append([1]+[float(line_cur[i]) for i in range(21)])
        trainlabel.append(float(line_cur[21]))

    trainweight = class_func(trainset, trainlabel)

    errnu, tolnum= 0, 0
    for line in f_test.readlines():
        line_cur = line.strip().split('	')
        pred_class = logistic_classify([1]+[float(line_cur[i]) for i in range(21)], trainweight)
        read_class = float(line_cur[21])
        if pred_class == read_class:
            #print "class succ"
            pass
        else:
            errnu += 1
            #print "class fail, read_class=%d, pred_class=%d" %(read_class, pred_class)
        tolnum += 1

    print "totol num=%d, fail num = %d, rate = %f" % (tolnum, errnu, float(errnu)/tolnum)

if __name__ == '__main__':
    test(grad_ascent)
    test(rand_grad_ascent)

来自：http://blackproof.iteye.com/blog/2064084

仅供学习参考