机器学习 -- 梯度下降法（Ⅶ 梯度下降法向量化计算实际数据）

本次我们利用之前向量化的梯度下降法进行波士顿房价预测。

1. 线性回归模型：里面包含正规化方程训练模型和梯度下降法训练模型的方式。

import numpy as np
from sklearn.metrics import r2_score


class LinearRegression:

    # 初始化LinearRegression模型
    def __init__(self):
        self.coef_ = None
        self.interception_ = None
        # theta为私有变量
        self._theta = None

    # 使用正规化方程
    def fit_normal(self, X_train, y_train):
        assert X_train.shape[0] == y_train.shape[0], \
            "X_train和y_train大小必须一致"

        X_b = np.hstack([np.ones((len(X_train), 1)), X_train])
        self._theta = np.linalg.inv(X_b.T.dot(X_b)).dot(X_b.T).dot(y_train)

        self.interception_ = self._theta[0]
        self.coef_ = self._theta[1:]

        return self

    def predict(self, X_predict):
        assert self.interception_ is not None and self.coef_ is not None, \
            "必须先拟合才能进行预测！"
        assert X_predict.shape[1] == len(self.coef_), \
            "预测值的特征数必须和训练集个数一致！"
        X_b = np.hstack([np.ones((len(X_predict), 1)), X_predict])
        return X_b.dot(self._theta)

    # 根据测试数据集X_test和y_test确定当前模型的准确度
    def score(self, X_test, y_test):
        y_predict = self.predict(X_test)
        return r2_score(y_test, y_predict)

    def fit_gd(self, X_train, y_train, eta=0.01, n_iters=1e4):
        """
        根据训练数据集X_train, y_train, 使用梯度下降法训练Linear Regression模型
        :param X_train:
        :param y_train:
        :param eta:
        :param n_iters:
        :return:
        """
        assert X_train.shape[0] == y_train.shape[0], "训练集X和y必须大小一致"

        def J(theta, X_b, y):
            """
            求代价函数
            :param theta: 参数值
            :param X_b:
            :param y:
            :return: 代价函数值
            """
            try:
                return np.sum((y - X_b.dot(theta)) ** 2) / len(y)
            except:
                return float('inf')

        def dJ(theta, X_b, y):
            return X_b.T.dot(X_b.dot(theta) - y) * 2. / len(X_b)

        def gradient_descent(X_b, y, initial_theta, eta, n_iters=1e4, epsilon=1e-8):
            theta = initial_theta
            i_iter = 0

            while i_iter < n_iters:
                gradient = dJ(theta, X_b, y)
                last_theta = theta
                theta = theta - eta * gradient
                if (abs(J(theta, X_b, y) - J(last_theta, X_b, y)) < epsilon):
                    break
                i_iter += 1
            return theta

        X_b = np.hstack([np.ones((len(X_train), 1)), X_train])  # 为原先的X添加一列,(())
        init_theta = np.zeros(X_b.shape[1])

        self._theta = gradient_descent(X_b, y_train, init_theta, eta, n_iters)
        self.interception_ = self._theta[0]
        self.coef_ = self._theta[1:]

        return self

2. 编写一个测试类用于测试。

（1）导入对应的包：

from sklearn import datasets
from demo1_knn.model_selection import train_test_split
from LinearRegression import LinearRegression
from sklearn.preprocessing import StandardScaler

（2）获取波士顿房价部分数据集：

def load_boston_data():
    """
    加载波士顿数据集并获取相应数据
    :return:
    """
    boston = datasets.load_boston()
    X = boston.data
    y = boston.target
    X = X[y < 50.0]
    y = y[y < 50.0]
    return X, y

（3）主函数：

if __name__ == '__main__':
    X, y = load_boston_data()
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_radio=0.2, seed=666)
    # 使用正规化方程训练模型
    reg = LinearRegression()
    reg.fit_normal(X_train, y_train)
    print(reg.score(X_test, y_test))  # 0.8129794056212729

    # 使用梯度下降法训练模型
    reg2 = LinearRegression()
    # reg2.fit_gd(X_train, y_train)
    reg2.fit_gd(X_train, y_train, eta=0.000001)
    print(reg2.score(X_test, y_test))  # 0.27586818724477224

    # 使用梯度下降法前进行数据归一化
    standardScaler = StandardScaler()
    standardScaler.fit(X_train)
    X_train_standard = standardScaler.transform(X_train)

    reg3 = LinearRegression()
    reg3.fit_gd(X_train_standard, y_train)
    X_test_standard = standardScaler.transform(X_test)
    print(reg3.score(X_test_standard, y_test))  # 0.8129873310487505

使用梯度下降法时，最好进行数据归一化。若所有数值不在一个维度上，会影响梯度计算的结果。reg3就是在梯度下降法之前使用了数据归一化处理。