python——回归案例

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
plt.rcParams['font.sans-serif'] = ['stxiHei']from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression
import seaborn as sns
from sklearn.metrics import r2_score, mean_squared_error




data = pd.read_csv(r'Statistics汽车销售数据.csv',encoding = 'utf-8')
data = data[data['公路里程数'].notna()]#只有1行缺失值，直接舍弃
data = data[['传统汽车销量','国内生产总值当季值(亿元)x1', '汽油价格（元/吨）x2',
       '人民币贷款基准利率%x3', '汽车总产量（万辆）x4', '公路里程数', '汽车整车股票指数', '消费者信心指数']]
data.head()



#数据的相关关系
cormatrix = data.corr() 
cormatrix *= np.tri(*cormatrix.values.shape,k=-1).T
cormatrix



#计算相关系数
corr_all = data.corr() 

#创建分布
mask = np.zeros_like(corr_all,dtype = np.bool) #
mask[np.triu_indices_from(mask)] = True#创建firue,画出heatmap
g,ax = plt.subplots(figsize = (9,7))
sns.heatmap(corr_all,mask = mask,square = True, linewidths = .5,ax = ax ,cmap = 'BuPu')
plt.title('Correlation of Feactures')
plt.show()


#特征选取
X = data[['国内生产总值当季值(亿元)x1', '汽油价格（元/吨）x2', '人民币贷款基准利率%x3','公路里程数', '汽车整车股票指数', '消费者信心指数']]
y = data['传统汽车销量']


plt.figure()for i in range(len(X.columns)):
    plt.scatter(X.iloc[:,i],
                y,
                color = np.array(plt.cm.tab10(i/len(X.columns))),
               label = X.columns[i])
    plt.legend()
    plt.show()


X = data[['国内生产总值当季值(亿元)x1', '汽油价格（元/吨）x2','公路里程数', '汽车整车股票指数', '消费者信心指数']]
y = data['传统汽车销量']
X.head()


#划分训练和测试数据集，为后续具有可重复性，设定随机种子random_state = 666
X_train, X_test, y_train, y_test = train_test_split(X,y,random_state = 666)



#模型
lin_reg0 = LinearRegression()
lin_reg0.fit(X_train,y_train)
y_predict = lin_reg0.predict(X_test)print('r2_score:' + str(r2_score(y_test,y_predict)))print('MSE:' + str(mean_squared_error(y_test,y_predict)))

#定义个绘制参数的函数，绘制残差图：
def plot_risiduals(model,X_train,X_test,y_train,y_test):
    fig,ax = plt.subplots(figsize = (7,4))

    risiduals_train = model.predict(X_train) - y_train
    ax.scatter(y_train,risiduals_train,label = 'Train',color = 'r')

    risiduals_test = model.predict(X_test) - y_test
    ax.scatter(y_test,risiduals_test,label = 'Test',color = 'k') 

    plt.axhline(y=0.0, c="b", ls="--", lw=2)
    plt.title('Risiduals')
    plt.xlabel('True')
    plt.ylabel('Risiduals')
plot_risiduals(lin_reg0,X_train,X_test,y_train,y_test)

#调整特征数量
X_train = X_train[['国内生产总值当季值(亿元)x1', '汽车整车股票指数', '消费者信心指数']]
X_test = X_test[['国内生产总值当季值(亿元)x1', '汽车整车股票指数', '消费者信心指数']]
lin_reg1 = LinearRegression()
lin_reg1.fit(X_train,y_train)
y_predict = lin_reg1.predict(X_test)print('r2_score:' + str(r2_score(y_test,y_predict)))print('MSE:' + str(mean_squared_error(y_test,y_predict)))
r2_score:0.9167941097031658
MSE:3878.5666590026112
plot_risiduals(lin_reg1,X_train,X_test,y_train,y_test)


#预测
True_Predict = {'True':y_test,'Predict':y_predict,'Risiduals':y_test-y_predict}
pd.DataFrame(True_Predict)

转自：https://mp.weixin.qq.com/s/o3TIX_7t2nsc6z-J5Lo-XQ