import numpy as np
import pandas as pd
import torch
from torch import nn
from multiprocessing import Pool
import matplotlib.pyplot as plt
import os
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
DAYS_FOR_TRAIN = 10
torch.manual_seed(42)
class LSTM_Regression(nn.Module):

    def __init__(self, input_size, hidden_size, output_size=1, num_layers=2):
        super().__init__()

        self.lstm = nn.LSTM(input_size, hidden_size, num_layers)
        self.fc = nn.Linear(hidden_size, output_size)

    def forward(self, _x):
        x, _ = self.lstm(_x)  # _x is input, size (seq_len, batch, input_size)
        s, b, h = x.shape  # x is output, size (seq_len, batch, hidden_size)
        x = x.view(s * b, h)
        x = self.fc(x)
        x = x.view(s, b, -1)  # 把形状改回来
        return x


def create_dataset(data, days_for_train=5) -> (np.array, np.array):
    dataset_x, dataset_y = [], []
    for i in range(len(data) - days_for_train):
        _x = data[i:(i + days_for_train)]
        dataset_x.append(_x)
        dataset_y.append(data[i + days_for_train])
    return (np.array(dataset_x), np.array(dataset_y))

def normal(nd):
    high = nd.describe()['75%'] + 1.5*(nd.describe()['75%']-nd.describe()['25%'])
    low = nd.describe()['25%'] - 1.5*(nd.describe()['75%']-nd.describe()['25%'])
    return nd[(nd<high)&(nd>low)]

def run(file_dir,excel):
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

    data = pd.read_excel(os.path.join(file_dir,excel), sheet_name=0,index_col=' stat_date ')

    data.columns = data.columns.map(lambda x: x.strip())

    data.index = pd.to_datetime(data.index,format='%Y%m%d')
    data.sort_index(inplace=True)
    print(data.head())
    data = data.loc['2021-01':'2023-09']

    data.drop(columns=[i for i in data.columns if (data[i] == 0).sum() / len(data) >= 0.5], inplace=True)  # 去除0值列
    print('len(data):', len(data))
    for industry in data.columns:
        df = data[industry]
        df = df[df.values != 0]  # 去除0值行
        df = normal(df)
        df = df.astype('float32').values  # 转换数据类型


        # 标准化到0~1
        max_value = np.max(df)
        min_value = np.min(df)
        df = (df - min_value) / (max_value - min_value)

        dataset_x, dataset_y = create_dataset(df, DAYS_FOR_TRAIN)
        print('len(dataset_x:)', len(dataset_x))

        # 划分训练集和测试集
        train_size = len(dataset_x) - 3
        train_x = dataset_x[:train_size]
        train_y = dataset_y[:train_size]

        # 将数据改变形状，RNN 读入的数据维度是 (seq_size, batch_size, feature_size)
        train_x = train_x.reshape(-1, 1, DAYS_FOR_TRAIN)
        train_y = train_y.reshape(-1, 1, 1)

        # 转为pytorch的tensor对象
        train_x = torch.from_numpy(train_x).to(device)
        train_y = torch.from_numpy(train_y).to(device)

        model = LSTM_Regression(DAYS_FOR_TRAIN, 32, output_size=1, num_layers=2).to(device)  # 导入模型并设置模型的参数输入输出层、隐藏层等


        train_loss = []
        loss_function = nn.MSELoss()
        optimizer = torch.optim.Adam(model.parameters(), lr=0.005, betas=(0.9, 0.999), eps=1e-08, weight_decay=0)
        for i in range(2000):
            out = model(train_x)
            loss = loss_function(out, train_y)
            loss.backward()
            optimizer.step()
            optimizer.zero_grad()
            train_loss.append(loss.item())
            # print(loss)
        # 保存模型
        # torch.save(model.state_dict(),save_filename)
        # torch.save(model.state_dict(),os.path.join(model_save_dir,model_file))

        # for test
        model = model.eval()  # 转换成测试模式
        # model.load_state_dict(torch.load(os.path.join(model_save_dir,model_file)))  # 读取参数
        dataset_x = dataset_x.reshape(-1, 1, DAYS_FOR_TRAIN)  # (seq_size, batch_size, feature_size)
        dataset_x = torch.from_numpy(dataset_x).to(device)

        pred_test = model(dataset_x)  # 全量训练集
        # 模型输出 (seq_size, batch_size, output_size)
        pred_test = pred_test.view(-1)
        pred_test = np.concatenate((np.zeros(DAYS_FOR_TRAIN), pred_test.cpu().detach().numpy()))

        # plt.plot(pred_test, 'r', label='prediction')
        # plt.plot(df, 'b', label='real')
        # plt.plot((train_size, train_size), (0, 1), 'g--')  # 分割线 左边是训练数据 右边是测试数据的输出
        # plt.legend(loc='best')
        # plt.show()


        # 创建测试集
        result_list = []
        # 以x为基础实际数据，滚动预测未来3天
        x = torch.from_numpy(df[-14:-4]).to(device)
        for i in range(3):
            next_1_8 = x[1:]
            next_9 = model(x.reshape(-1,1,DAYS_FOR_TRAIN))
            # print(next_9,next_1_8)
            x = torch.concatenate((next_1_8, next_9.view(-1)))
            result_list.append(next_9.view(-1).item())


        # 反归一化
        pred = np.array(result_list) * (max_value - min_value) + min_value

        df = df * (max_value - min_value) + min_value


        # 打印指标
        # print(abs(pred - df[-3:]).mean() / df[-3:].mean())
        result_eight = pd.DataFrame({'pred': np.round(pred,1),'real': df[-3:]})
        target = (result_eight['pred'].sum() - result_eight['real'].sum()) / df[-31:].sum()
        result_eight['loss_rate'] = round(target, 5)
        print(target)
        print(result_eight)

        result_eight.to_csv(f'./月底预测结果/9月{excel[:2]}.txt', sep='\t', mode='a')
        with open(fr'./偏差/9月底偏差率.txt', 'a', encoding='utf-8') as f:
            f.write(f'{excel[:2]}{industry}:{round(target, 5)}\n')

if __name__ == '__main__':
    file_dir = r'C:\Users\user\PycharmProjects\pytorch2\杭州日电量\浙江所有地市133行业数据'

    # run(file_dir,'丽水133行业数据（全）.xlsx')
    p = Pool(4)
    for excel in os.listdir(file_dir):
        p.apply_async(func=run,args=(file_dir,excel))
    p.close()
    p.join()