逻辑回归可视化——绘制决策边界

具体代码如下，现在的代码是基于之前的逻辑回归Sigmoid函数的得到的结果最优参数，这是之前博客的地址：https://blog.csdn.net/qq_41938259/article/details/104139237

使用的数据集也在这个链接里给出了，

以下附上代码，代码使用了psutil来显示实时的内存使用率，和3百万次迭代的for循环的运算时间。

import os
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import psutil
import datetime


# 定义sigmoid函数
def sigmoid(x):
    return 1.0/(1.0 + np.exp(-x))


# 逻辑回归计算参数的核心
# 会涉及numpy矩阵运算
def logicRegression(data, label):
    dataMatrix = data.to_numpy()
    labelMat = label.to_numpy()
    m, n = dataMatrix.shape
    alpha = 0.001
    weights = np.ones((n, 1))
    startTime = datetime.datetime.now()
    for cycle in range(3000000):
        mem = psutil.virtual_memory()
        print('内存利用率：{}\n一共内存：{}MB\n空闲内存：{}MB\n'.format(mem.percent, mem.total/(2**20), mem.free/(2**20)))
        vector = sigmoid(dataMatrix.dot(weights))
        error = labelMat - vector
        weights = weights + alpha * (dataMatrix.T).dot(error)
    endTime = datetime.datetime.now()
    print('for循环运行时间：{}秒'.format((startTime-endTime).seconds))
    return weights


# 可视化模型x1、y1是faster的距离（副作用）和收入/生活费（正作用）
# x2、y2是lower的距离（副作用）和收入/生活费（正作用）
def visualize_model(x1, y1, x2, y2):
    fig = plt.figure(figsize=(6, 6), dpi=80)
    ax = fig.add_subplot(111)
    ax.set_xlabel("$distance$")
    ax.set_xticks(range(0, 3000, 500))
    ax.set_ylabel("$money$")
    ax.set_yticks(range(0, 4000, 500))
    ax.scatter(x1, y1, color="b", alpha=0.4)
    ax.scatter(x2, y2, color="r", alpha=0.4)
    plt.legend(shadow=True)
    plt.show()


def draw(result, data, label):
    data = np.array(data)
    label = np.array(label)
    m,n = data.shape
    x1 = []
    y1 = []
    x2 = []
    y2 = []
    for i in range(m):
        if int(label[i]) == 1:
            x1.append(data[i, 1])
            y1.append(data[i, 2])
        else:
            x2.append(data[i, 1])
            y2.append(data[i, 2])
    fig = plt.figure(figsize=(8, 8), dpi=80)
    ax = fig.add_subplot(111)
    ax.scatter(x1, y1, color="b", alpha=0.4)
    ax.scatter(x2, y2, color="r", alpha=0.4)
    ax.set_xlabel("$distance$")
    ax.set_xticks(range(0, 3000, 500))
    ax.set_ylabel("$money$")
    ax.set_yticks(range(0, 4000, 500))
    x = range(0, 3000, 500)
    y = (result[0]+result[1]*x)/result[2]
    ax.plot(x, y)
    plt.show()


if __name__ == '__main__':
    # 打开文件操作
    os.chdir('D:\\')
    # 读取实验集
    data = pd.read_excel('附件1.xlsx', sep=',')
    result = data['III']
    distance = data['II']
    money = data['VI']
    X = data['IV']
    Y = data['X']
    mistake = data['V']
    test1 = pd.DataFrame({'result': result, 'distance': distance, 'money': money, 'mistake': mistake})

    # 删去因为取票，而不得买错票的
    # faster是买高铁票的人，而且是买对的
    # lower是买普快的人，也是买对的
    test1 = test1[(test1.mistake == 0)]
    faster = test1[(test1.result == 1)]
    lower = test1[test1.result == 0]

    # 整理数据
    faster = pd.DataFrame({'distance': faster['distance'], 'money': faster['money']})
    lower = pd.DataFrame({'distance': lower['distance'], 'money': lower['money']})
    # 丢弃有误数据
    lower = lower.drop(index=129)

    # 可视化步骤，红单点标签值为0，蓝点为1
    # visualize_model(faster['distance'], faster['money'], lower['distance'], lower['money'])

    # 准备逻辑回归的数据集
    m, n = test1.shape
    datas = pd.DataFrame({'X0': np.array([1]*m), 'X1': test1['distance'], 'X2': test1['money']})
    labels = pd.DataFrame({'label': test1['result']})
    # 运行伙计回归并打印结果
    result = logicRegression(datas, labels)
    print(result)
    draw(result, datas, labels)

这里有输出结果：

                        

 

---

END