线性回归算法

简单线性回归

• 简单线性回归包含一个自变量(x)和一个因变量(y)，如果包含两个以上的自变量，则称作多元回归分析(multiple regression)，被用来描述因变量(y)和自变量(X)以及偏差(error)之间关系的方程叫做回归模型
  • 公式：
    $$\sum _{m=0}^m(y^{(i)}-a\ast x^{(i)}-b{)}^2\text{，一元线性回归算法公式}$$
  • 原理:找到a和b使得上述公式的值尽可能的小，经过最小二乘法求解可以计算出a值是：
    $$a=\frac{{\sum }_{m=0}^m(x^{(i)}-\overline{x})\ast (y^{(i)}-\bar{y})}{{\sum }_{m=0}^m(x^{(i)}-\bar{x}{)}^2}\text{，a的值}$$
    $$b=\bar{y}-a\ast \bar{x}\text{，b的值}$$
  • 使用代码实现如下：这种方式实现的代码时间复杂度很大O(n^3)不建议使用

class SimpleLinearRegression1:
    def __init__(self):
        self.a_ = None
        self.b_ = None

    def fit(self, x_train, y_train):
        """根据训练数据集x_train,y_train训练Simple Linear Regression模型"""
        if x_train.ndim != 1:
            raise Exception("简单回归的维度只能是一维")
        if len(x_train) != len(y_train):
            raise Exception("x_train的行数需要和y_train的行数相同")
        """计算训练集平均值"""
        x_mean = np.mean(x_train)
        """计算训练集平均值"""
        y_mean = np.mean(y_train)

        num = 0.0  # 分子
        d = 0.0  # 分母
        """循环得出x，y的值"""
        for x, y in zip(x_train, y_train):
            """利用最小二乘法计算出a、b的值"""
            num += (x - x_mean) * (y - y_mean)
            d += (x - x_mean) ** 2
        self.a_ = num / d
        self.b_ = y_mean - self.a_ * x_mean
        return self

    def predict(self, x_predict):
        """x_predict是待预测的数据集，输入这个x_predict会得到一个预测值"""
        if x_predict.ndim != 1:
            raise Exception("简单回归的维度只能是一维")
        if self.a_ is None and self.b_ is None:
            raise Exception("预测之前需要先拟合")
        res = []
        for x in x_predict:
            res.append(self._predict(x))
            """封装为np.array格式数据"""
        return np.array(res)

    def _predict(self, x_single):
        """转化为函数y=a*x+b"""
        return self.a_ * x_single + self.b_

• 第二方式实现
  $$a=\frac{(x^{(i)}-\overline{x})\cdot (y^{(i)}-\bar{y})}{(x^{(i)}-\bar{x}{)}^2}\text{，a的值}$$
  $$b=\bar{y}-a\ast \bar{x}\text{，b的值}$$
  • 代码如下：

class SimpleLinearRegression2:
    def __init__(self):
        self.a_ = None
        self.b_ = None

    def fit(self, x_train, y_train):
        """根据训练数据集x_train,y_train训练Simple Linear Regression模型"""
        if x_train.ndim != 1:
            raise Exception("简单回归的维度只能是一维")
        if len(x_train) != len(y_train):
            raise Exception("x_train的行数需要和y_train的行数相同")
        x_mean = np.mean(x_train)
        y_mean = np.mean(y_train)

        num = 0.0  # 分子
        d = 0.0  # 分母
        num = (x_train - x_mean).dot(y_train - y_mean)
        d = (x_train - x_mean).dot(x_train - x_mean)
        self.a_ = num / d
        self.b_ = y_mean - self.a_ * x_mean
        return self

    def predict(self, x_predict):
        """x_predict是待预测的数据集，输入这个x_predict会得到一个预测值"""
        if x_predict.ndim != 1:
            raise Exception("简单回归的维度只能是一维")
        if self.a_ is None and self.b_ is None:
            raise Exception("预测之前需要先拟合")
        res = []
        for x in x_predict:
            res.append(self._predict(x))
        return np.array(res)

    def score(self, x_test, y_test):
        y_predict = self.predict(x_test)
        return r2_score(y_test, y_predict)

    def _predict(self, x_single):
        return self.a_ * x_single + self.b_

• 评价一盒回归算法的好坏可以利用MSE(均方误差)、RMSE(均方根误差)、MAE(平均绝对误差)、R2(决定系数)实现代码如下

def mean_squared_error(y_true, y_predict):
    if len(y_true) != len(y_predict):
        raise Exception("y_ture的长度应该和y_predict相同")
    return np.sum((y_true - y_predict) ** 2) / len(y_true)


def root_mean_squared_error(y_true, y_predict):
    """计算y_true和y_predict之间的RMSE"""
    return sqrt(mean_squared_error(y_true, y_predict))


def mean_absolute_error(y_true, y_predict):
    """计算y_true和y_predict之间的MAE"""
    return np.sum(np.absolute(y_true - y_predict)) / len(y_true)


def r2_score(y_true, y_predict):
    return 1 - mean_squared_error(y_true, y_predict) / np.var(y_true)

多元线性回归:

• 公式
  $$y={\theta }_0+{\theta }_1{x}_1+{\theta }_2{x}_2+{\theta }_3{x}_3+......+{\theta }_n{x}_n\text{，y的值}$$
• 分解这个函数可以得到：
  $${\hat{y}}^{(i)}=x_0+x_1+x_2+x_3+......+x_n$$
  $${\theta }_{=}({\theta }_0+{\theta }_1+{\theta }_2+{\theta }_3+......+{\theta }_n{)}^T$$根据向量化运算可以得到$${\theta }_{=}(x_b\cdot x_b{)}^T\cdot x_b^T\cdot y$$ 函数的截距就是$${\theta }_0$$系数是$${\theta }_1......{\theta }_n$$

• y值，就是数学函数输出的值是
  $$y=x_b\cdot {\theta }_{}$$ $${\theta }_{=}(x_b\cdot x_b{)}^T\cdot x_b^T\cdot y$$
  np.hstack():在水平方向上平铺
  X_b = np.hstack([np.ones((len(X_train), 1)), X_train])
• 多元线性回归代码如下:

# -*- coding: utf-8 -*-
# @Time    : 2021/7/30 10:56
# @Email   : 2635681517@qq.co
import numpy as np

from metrics import r2_score


class LinearRegression:
    def __init__(self):
        self.coef_ = None
        self.intercept_ = None
        self.theta_ = None

    def fit_normal(self, X_train, y_train):
        """根据训练数据集X_train, y_train训练Linear Regression模型"""
        if X_train.shape[0] != y_train.shape[0]:
            raise Exception("X_train的行数需要和y_train的行数相同")
        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.intercept_ = self.theta_[0]
        self.coef_ = self.theta_[1:]
        return self

    def predict(self, X_predict):
        """给定待预测数据集X_predict，返回表示X_predict的结果向量"""
        assert self.intercept_ is not None and self.coef_ is not None, \
            "must fit before predict!"
        assert X_predict.shape[1] == len(self.coef_), \
            "the feature number of X_predict must be equal to X_train"

        X_b = np.hstack([np.ones((len(X_predict), 1)), X_predict])
        return X_b.dot(self.theta_)

    def score(self, X_test, y_test):
        """根据测试数据集 X_test 和 y_test 确定当前模型的准确度"""

        y_predict = self.predict(X_test)
        return r2_score(y_test, y_predict)

    def __repr__(self):
        return "LinearRegression()"