# 开发者：董书雄
# 开发时间:2022-12-01 22:19
# 开发内容:{}
# 开发者：董书雄
# 开发时间:2022-11-13 23:33
# 开发内容:{}
import numpy as np
from sklearn.model_selection import train_test_split
import pandas as pd
import joblib
from sklearn.gaussian_process import GaussianProcessRegressor
from sklearn import metrics, ensemble
from catboost import CatBoostRegressor # 调用模型
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.gaussian_process.kernels import RBF, ConstantKernel as C
pd.set_option('display.max_columns', None) # 结果显示所有列
pd.set_option('display.max_rows', None) # 结果显示所行行
# 1：TGAN合成数据的训练效果
# data1 = pd.read_csv('traindata928.csv')
# data2 = pd.read_csv('testdata.csv')
# 2：纯原数据232个的训练效果
data2 = pd.read_csv('testdata.csv')
data1, data2 = train_test_split(data2, test_size=0.3)
print(data1)
print(data2)
# 记得把测试和训练数据集的特征第一行特征整一样！
# print(data1,type(data1))
x_train = data1.iloc[:, 0:9]
y_train = data1.iloc[:, 9]
x_test = data2.iloc[:, 0:9]
y_test = data2.iloc[:, 9]
X = pd.concat([x_train, x_test])
Y = pd.concat([y_train, y_test])
col = data1.columns[:-1]
# print(X)
# print(x_train)
# 训练模型
kernel = C(10, (0.1, 0.1)) * RBF(0.5, (1e-4, 10))
rfmodel = GaussianProcessRegressor(kernel=kernel, n_restarts_optimizer=5, alpha=0.05
) #
rfmodel.fit(x_train, y_train)
# 模型预测
joblib.dump(rfmodel, 'rffinal.model')
rfmodel = joblib.load('rffinal.model')
rf_predict_test = rfmodel.predict(x_test)
rf_predict_train = rfmodel.predict(x_train)
rf_predict_all = rfmodel.predict(X)
rf_error = rf_predict_test - y_test
# 计算性能指标
rf_R2_rtrain = metrics.r2_score(y_train, rf_predict_train)
rf_R2_rtest = metrics.r2_score(y_test, rf_predict_test)
rf_R2_rall = metrics.r2_score(Y, rf_predict_all)
rf_RMSE_train = metrics.mean_squared_error(y_train, rf_predict_train) ** 0.5
rf_RMSE_test = metrics.mean_squared_error(y_test, rf_predict_test) ** 0.5
rf_RMSE_all = metrics.mean_squared_error(Y, rf_predict_all) ** 0.5
rf_MAPE_train = metrics.mean_absolute_percentage_error(y_train, rf_predict_train)
rf_MAPE_test = metrics.mean_absolute_percentage_error(y_test, rf_predict_test)
rf_MAPE_all = metrics.mean_absolute_percentage_error(Y, rf_predict_all)
rf_performance = (rf_R2_rtrain, rf_R2_rtest, rf_R2_rall, rf_RMSE_train,
rf_RMSE_test, rf_RMSE_all, rf_MAPE_train, rf_MAPE_test, rf_MAPE_all)
# 绘图数据
data_array = np.array(rf_predict_test)
rf_predict_test = data_array.tolist()
data_array = np.array(rf_predict_all)
rf_predict_all= data_array.tolist()
print('y_test=', y_test)
print('rf_predict_test: %0.3f', rf_predict_test, type(rf_predict_test))
print('Y=', Y)
print('rf_predict_all=', rf_predict_all)
print('rf_R2_rtrain=', rf_R2_rtrain)
print('rf_R2_rtest=', rf_R2_rtest)
print('rf_R2_rall=', rf_R2_rall)
print('rf_RMSE_train=', rf_RMSE_train)
print('rf_RMSE_test=', rf_RMSE_test)
print('rf_RMSE_all=', rf_RMSE_all)
print('rf_MAPE_train=', rf_MAPE_train)
print('rf_MAPE_test=', rf_MAPE_test)
print('rf_MAPE_all=', rf_MAPE_all)
#plt.scatter(Y, rf_predict_all, )
#plt.show()
