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[(ndlow)] device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') data = pd.read_excel(r'C:\Users\鸽子\Desktop\浙江133行业日电量数据.xlsx', 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)) df_result = pd.DataFrame({'预测值':[],'实际值':[],'偏差率':[],'行业':[]}) 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({'预测值': np.round(pred,1),'实际值': df[-3:]}) target = (result_eight['预测值'].sum() - result_eight['实际值'].sum()) / df[-31:].sum() result_eight['偏差率'] = round(target, 5) result_eight['行业'] = industry df_result = pd.concat((df_result,result_eight)) print(df_result) # result_eight.to_csv(f'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')