python+sklearn 机器学习代码备忘

import sklearn
from sklearn.model_selection import train_test_split 
from sklearn.linear_model import LinearRegression  
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn import preprocessing
import csv
import numpy as np
from sklearn.inspection import partial_dependence
from sklearn.inspection import plot_partial_dependence
from sklearn.inspection import partial_dependence
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import cross_val_score
import shap
from sklearn.preprocessing import OneHotEncoder
from matplotlib.pyplot import MultipleLocator
from sklearn.metrics import r2_score


#读取原表格
pd_data=pd.read_csv('QYD.csv')

#提取特征与标签，此处X_origin输入特征，Y_origin输入因变量
X_origin = pd_data.loc[:, ('BTSM','SVF','GVI','ROAD','POI','JZMD','NDVI','DLMD','POPU','FLOOR')]
y_origin = pd_data.loc[:, 'temp']

X, y = sklearn.utils.shuffle(X_origin, y_origin)

#划分测试集和训练集
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=532)#选择20%为测试集
print('训练集测试及参数:')
print('X_train.shape={}\n y_train.shape ={}\n X_test.shape={}\n,  y_test.shape={}'.format(X_train.shape,y_train.shape,X_test.shape,y_test.shape))

#定义训练模型
rfreg = RandomForestRegressor(n_estimators=161, max_depth=19, random_state=90)

#训练
model = rfreg.fit(X_train, y_train)

#输出模型超参数
print('模型参数:')
print(model)

# 训练后模型截距
print('模型特征重要性:')
print(rfreg.feature_importances_)

# 训练后模型权重
score = cross_val_score(rfreg, X_train, y_train, cv=10).mean()
print(score)


#对测试集进行训练
y_pred = rfreg.predict(X_test)

#初始化误差参数
sum_mean = 0

#计算均方误差
for i in range(len(y_pred)):
    sum_mean += (y_pred[i] - y_test.values[i]) ** 2
sum_erro = np.sqrt(sum_mean /len(y_pred))

#打印均方误差
print ("RMSE by hand:", sum_erro)
print("R2:", rfreg.score(X_test,y_test))


#返回十次交叉验证值,此处返回负的MSE，转为正值
cross_score=cross_val_score(rfreg, X, y, cv=10,scoring = 'neg_mean_squared_error')
print(pow(-cross_score,0.5))

# 做验证集的原始数据-预测数据曲线
plt.figure()
plt.plot(range(len(y_pred)), y_pred, 'r', label="predict")
plt.plot(range(len(y_pred)), y_test, 'b', label="test",)
plt.legend(loc="upper right")  # 显示图中的标签
plt.xlabel("the number of sales")
plt.ylabel('value of sales')
plt.show()
plt.close()


#绘制特征的相关性热力图
hot = X.corr()
plt.subplots(figsize = (4,4))
sns.heatmap(hot,annot = True,vmax = 1,square = True,cmap = "Blues")
plt.show()
plt.close()



#绘制部分依赖图，输入特征
features = ['BTSM','SVF','GVI','ROAD','POI','JZMD','NDVI','DLMD','POPU','FLOOR']
plot_partial_dependence(rfreg, X_train, features,n_jobs=3, grid_resolution=20, method='brute')#,line_kw={"color": "black","lw":0.8}，line_kw是传给plot的关键字字典
fig = plt.gcf()
fig.subplots_adjust(hspace=0.3)
plt.show()
plt.close()



#shap可解释机器学习
#创建explainer，输入特征
cols = ['BTSM','SVF','GVI','ROAD','POI','JZMD','NDVI','DLMD','POPU','FLOOR']
explainer = shap.TreeExplainer(rfreg)

#numpy.array数组
shap_values = explainer.shap_values(pd_data[cols])

#shap.Explanation对象
shap_values2 = explainer(pd_data[cols])

print(shap_values.shape)

#绘制Summary Plot
shap.summary_plot(shap_values, pd_data[cols])

#绘制heatmap
shap.plots.heatmap(shap_values2)