机器学习之梯度下降算法模型

代码如下：

from sklearn.metrics import mean_absolute_error,mean_squared_error,r2_score
import numpy as np
import matplotlib.pyplot as plt


# 归一化函数
def featureNormalize(X):
    mu = X.mean(0)
    std = np.std(X)
    X = (X - mu) / std
    index = np.ones((X.shape[0], 1))
    X = np.hstack((X, index))
    return X


# 读数据
def getData():
    data = np.loadtxt(r"D:\Pycharm\code\MY_machine_learning\mydata\aqi2.csv", delimiter=",", skiprows=1,
                      dtype=np.float32)
    X = data[:, 1:]  # 特征集 二维矩阵
    Y = data[:, 0]  # 目标（标签label）集
    Y = Y.reshape(-1, 1)  # 将Y转为1列二维矩阵，行数由列数自动计算
    # 特征归一化（特征缩放）
    X = featureNormalize(X)
    return X, Y


# 计算损失函数
def lossFunction(X, Y, theta):
    m = X.shape[0]  # 样本个数
    loss = sum((np.dot(X, theta) - Y) ** 2) / (2 * m)
    return loss


# 实现批量梯度下降算法
def batchGradientDecent(X, Y, theta, alpha, num_iters):
    m = X.shape[0];  # 样本个数
    loss_all = [];  # 定义列表
    for i in range(num_iters):
        theta = theta - alpha * np.dot(X.T, (np.dot(X, theta) - Y)) / m
        loss = lossFunction(X, Y, theta)
        loss_all.append(loss)
        print("第{}次的损失值为:{}".format(i + 1, loss))
    return theta, loss_all


if __name__ == "__main__":
    # 1. 读数据
    X, Y = getData()

    # 2. 初始化参数theta的值
    theta = np.ones((X.shape[1], 1))

    # 3.初始化迭代次数
    num_iters = 500
    alpha = 0.01

    # 4.调用梯度下降算法迭代更新参数
    theta, loss_all = batchGradientDecent(X, Y, theta, alpha, num_iters)
    print(theta)

    # 5.预测
    y_predict = np.dot(X, theta)
    print('打印预测值：\n', y_predict)
    # 6. 可视化数据
    # plt.scatter(np.arange(100), Y[:100], c="red")
    # plt.scatter(np.arange(100), y_predict[:100], c="green")

    # 绘制损失函数与迭代次数的关系曲线
    plt.plot(np.arange(num_iters), loss_all, c='r')
    plt.show()

    print("mae:", mean_absolute_error(Y, y_predict))
    print("mse:", mean_squared_error(Y, y_predict))
    print("r2:", r2_score(Y, y_predict))

运行结果如下：

 附aqi2.csv网盘提取

百度网盘 请输入提取码

提取码：uuyp