# '''
# 模型筛选
# '''

## 导入常用基本包
import os
import pandas as pd
import numpy as np
from PIL import Image
from model_saver import save_model

# 机器学习模型导入
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import cross_val_score,cross_val_predict
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error

## 导入常用辅助函数
from sklearn.model_selection import train_test_split
from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import cross_val_predict

## 导入数据处理函数
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import MinMaxScaler

## 导入评分函数
from sklearn.metrics import r2_score
from sklearn.metrics import mean_squared_error 
from sklearn.metrics import mean_absolute_error
from sklearn.metrics import accuracy_score
from sklearn.metrics import log_loss
from sklearn.metrics import roc_auc_score


# 导入数据
data=pd.read_excel('model_optimize\data\data_filt.xlsx')
x = data.iloc[:,1:10]
y = data.iloc[:,-1]

# 为 x 赋予列名
x.columns = [
    'organic_matter',        # OM g/kg
    'chloride',              # CL g/kg
    'cec',                   # CEC cmol/kg
    'h_concentration',       # H+ cmol/kg
    'hn',                    # HN mg/kg
    'al_concentration',      # Al3+ cmol/kg
    'free_alumina',          # Free alumina g/kg
    'free_iron',             # Free iron oxides g/kg
    'delta_ph'               # ΔpH
]

y.name = 'target_ph'

Xtrain, Xtest, Ytrain, Ytest=train_test_split(x, y, test_size=0.2)

# 筛选随机种子
score_5cv_all = []
for i in range(0, 200, 1):
    rfc =RandomForestRegressor(random_state=i)
    score_5cv =cross_val_score(rfc, Xtrain, Ytrain, cv=5).mean()
    score_5cv_all.append(score_5cv)
    pass
score_max_5cv = max(score_5cv_all)

random_state_5cv = range(0, 200)[score_5cv_all.index(max(score_5cv_all))] # 5cv最大得分对应的随机种子

print("最大5cv得分：{}".format(score_max_5cv),
      "random_5cv:{}".format(random_state_5cv))


# 筛选随机树数目
score_5cv_all = []
for i in range(1, 400, 1):
    rfc = RandomForestRegressor(n_estimators=i,
        random_state=random_state_5cv)
    score_5cv = cross_val_score(rfc, Xtrain, Ytrain, cv=5).mean()
    score_5cv_all.append(score_5cv)
    pass
score_max_5cv = max(score_5cv_all)
n_est_5cv = range(1,400)[score_5cv_all.index(score_max_5cv)]    # 5cv最大得分对应的树数目

print("最大5cv得分：{}".format(score_max_5cv),
      "n_est_5cv:{}".format(n_est_5cv))  # 5cv最大得分对应的树数目??
score_test_all = []


# 筛选最大深度
score_5cv_all = []
for i in range(1, 300, 1):
    rfc = RandomForestRegressor(n_estimators=n_est_5cv
                                , random_state=random_state_5cv
                                , max_depth=i)
    score_5cv = cross_val_score(rfc, Xtrain, Ytrain, cv=5).mean()
    score_5cv_all.append(score_5cv)
    pass
score_max_5cv = max(score_5cv_all)
max_depth_5cv = range(1,300)[score_5cv_all.index(score_max_5cv)]    
print(
      "最大5cv得分：{}".format(score_max_5cv),
      "max_depth_5cv:{}".format(max_depth_5cv))      

# 确定参数进行训练
rfc = RandomForestRegressor(n_estimators=n_est_5cv,random_state=random_state_5cv,max_depth=max_depth_5cv)
CV_score = cross_val_score(rfc, Xtrain, Ytrain, cv=5).mean()
CV_predictions = cross_val_predict(rfc, Xtrain, Ytrain, cv=5)
rmse1 = np.sqrt(mean_squared_error(Ytrain,CV_predictions))
regressor = rfc.fit(Xtrain, Ytrain)
test_predictions = regressor.predict(Xtest)
score_test = regressor.score(Xtest,Ytest)
rmse2 = np.sqrt(mean_squared_error(Ytest,test_predictions))
print("5cv:",CV_score)
print("rmse_5CV",rmse1)
print("test:",score_test)
print("rmse_test",rmse2)

# 保存训练好的模型
custom_path='model_optimize\pkl'         # 模型保存路径
prefix='rf_model_'          # 模型文件名前缀
save_model(rfc, custom_path, prefix)