线性回归Demo

1. 相关包导入

# -*- coding: utf-8 -*-
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error

all_names = ["gender" "age", "job", "industry", "rfm_label"]
feature_names = ["gender", "age"]

if __name__ == '__main__':
    linear_model()

2. 模型定义

def linear_model():
    data = pd.read_csv('../lr_data/train.csv', header=None, sep='\t').fillna(0)
    data.columns = all_names

    X = data.loc[:, feature_names]
    Y = data.loc[:, 'revolve_rate']
    X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size=0.3, random_state=0)

    transfer = StandardScaler()
    x_train = transfer.fit_transform(X_train)
    x_test = transfer.fit_transform(X_test)

    estimator = LinearRegression()   # LR线性回归
    # estimator = SGDRegressor(max_iter=1000)  # 梯度下降法
    # estimator = Ridge(alpha=1)  # 岭回归
    estimator.fit(x_train, y_train)

    # 5.模型评估
    y_predict = estimator.predict(x_test)
    print("预测值为:\n", y_predict)
    print("模型中的系数为:\n", estimator.coef_)
    print("模型中的偏置为:\n", estimator.intercept_)
    error = mean_squared_error(y_test, y_predict)
    print("误差为:\n", error)

3. R2_score

参考：r2_score使用方法_*Snowgrass*的博客-CSDN博客_r2_score

# python 内置方法
print(estimator.score(x_test, y_test))
# 自定义
def r2_score(y_true, y_predict):
    return 1 - mean_squared_error(y_true, y_predict) / np.var(y_true)

R2通俗地理解为使用均值作为误差基准，看预测误差是否大于或者小于均值基准误差。

R2_score=1:  样本中预测值和真实值完全相等，没有任何误差，表示回归分析中自变量对因变量的解释越好；

R2_score=0:  此时分子等于分母，样本的每项预测值都等于均值。