【机器学习】【逻辑回归】Python实现逻辑回归

1.逻辑回归python实现

# -*- coding: utf-8 -*-
"""
@author: 蔚蓝的天空Tom
"""

import numpy as np
import os
import matplotlib.pyplot as plt
from sklearn.datasets import make_blobs

#global variable
path = r'D:\tom\data'
train_set = 'train_lr.txt'
test_set = 'test_lr.txt'

#load data set from txt file
def load_data_set(path, file_name):
    data_list = []
    label_list = []
    f = open(os.path.join(path, file_name))
    for line in f.readlines():
        #feature1	feature2	 label
        text = line.strip().split()
        #data_m追加[feature0=1, feature1, feature2]
        #样本的实际特征值有2个，附加一个恒值为1的特征值feature0，以便运算
        #记作features_cnt = 1 + n_features
        data_list.append([1.0, float(text[0]), float(text[1])])
        #label_m追加label value
        label_list.append(int(text[2]))
    return data_list, label_list

#sigmoid function
def sigmoid(z):
    return 1/(1+np.exp(-1*z))
def sigmoid_for_m(Z):
    fin = 1
    fout = 1
    sigmoid_func = np.frompyfunc(sigmoid, fin, fout)
    return sigmoid_func(Z)

def trainLogisticRegression(data_list, label_list):
    #samples feature matrix, samples_cnt * features_cnt
    #其中samples_cnt = n_samples
    #其中features_cnt = 1 + n_features
    data_m = np.mat(data_list).astype(np.float64)
    #samples label matrix, samples_cnt * 1
    #其中samples_cnt = n_samples
    label_m = np.mat(label_list).T.astype(np.float64)
    #get samples features shape
    #其中m = samples_cnt = n_samples
    #其中n = features_cnt = 1 + n_samples
    m,n = np.shape(data_m)
    #初始化权重系数w，features_cnt * 1
    #其中features_cnt = 1 + n_features
    w = np.ones((n,1))
    #初始化学习效率learning rate, alpha
    alpha = 0.001
    #设置循环迭代计算的执行最大次数，loop_limit
    loop_limit = 1000
    #误差初始化
    error = 10000
    #开始迭代计算
    for i in range(loop_limit):
        predict = sigmoid_for_m(data_m * w)
        error = predict - label_m
        w = w - alpha * data_m.T * error
    return w

#train the test set
def train_test_set(weights, data):
    w = weights
    return w[0]/w[2] - w[1]/w[2]*data

#显示拟合图形
def plot_fit_line(weights, data, label):
    if type(weights).__name__ == 'ndarray':
        w = weights
    else:
        w = weights.getA()
    #test set
    test_data_list,test_label_list = make_blobs(n_samples=200, n_features=2, centers=2)
    
    #
    plt.figure(1)
    test_data_set = np.arange(-3, 3, 0.1)
    lr_test_label = -1 * w[0]/w[2] - w[1]/w[2]*test_data_set
    #lr_test_label = train_test_set(w, test_data_set)
    plt.plot(test_data_set, lr_test_label)
    
    negative = [] #否定的
    positive = [] #肯定的
    for i in range(len(label)):
        if 1 == label[i]:
            positive.append(data[i][1:3])
        else:
            negative.append(data[i][1:3])
    positive = np.array(positive)
    negative = np.array(negative)
    plt.scatter(positive[:,0], positive[:,1], c='red')
    plt.scatter(negative[:,0], negative[:,1], c='green')
    plt.show()

if __name__ == "__main__":
  data,label = load_data_set(path, train_set)
  data = np.array(data).astype(np.float64)
  label = [int(item) for item in label]
  weights = trainLogisticRegression(data,label)
  plot_fit_line(weights,data,label)

2.拟合结果

3.帮助文档

python实现逻辑回归的方法示例：http://www.jb51.net/article/112719.htm

scikit-learn文档：http://scikit-learn.org/stable/modules/linear_model.html#logistic-regression

逻辑回归在线交互服务器：http://www.vassarstats.net/logreg1.html#down

python编写Logistic逻辑回归：http://www.jb51.net/article/131182.htm

python实现逻辑回归的方法示例：http://www.jb51.net/article/112719.htm

(end)