import numpy as np import pandas as pd import torch from torch import nn import os import matplotlib.pyplot as plt from sklearn.preprocessing import MinMaxScaler from torch.utils.data import DataLoader,TensorDataset 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 inverse_transform_col(scaler,y,n_col): '''scaler是对包含多个feature的X拟合的,y对应其中一个feature,n_col为y在X中对应的列编号.返回y的反归一化结果''' y = y.copy() y -= scaler.min_[n_col] y /= scaler.scale_[n_col] return y device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') data = pd.read_excel(r'C:\python-project\pytorch3\入模数据\杭州数据.xlsx',index_col='dtdate') print(data.columns) data.columns = data.columns.map(lambda x: x.strip()) data.drop(columns='city_name',inplace=True) # 标准化到0~1 scaler = MinMaxScaler() df_normalized = pd.DataFrame(scaler.fit_transform(data), columns=data.columns,index=data.index).astype(float) # 划分训练集和测试集 x_all = df_normalized.drop(columns='售电量').loc['2021-1':'2023-9'] y_all = df_normalized['售电量'].loc['2021-1':'2023-9'] x_train = df_normalized.drop(columns='售电量').loc['2021-1':'2023-8'] y_train = df_normalized['售电量'].loc['2021-1':'2023-8'] eval = df_normalized.loc['2023-9'] x_eval = eval.drop(columns='售电量') y_eval = eval['售电量'] # 将数据改变形状,RNN 读入的数据维度是 (seq_size, batch_size, feature_size) x_train = np.array(x_train.values).reshape(-1, 1, 13) y_train = np.array(y_train.values).reshape(-1, 1, 1) # 转为pytorch的tensor对象 x_train = torch.from_numpy(x_train).to(device).type(torch.float32) y_train = torch.from_numpy(y_train).to(device).type(torch.float32) x_eval = torch.from_numpy(x_eval.values).to(device).type(torch.float32) model = LSTM_Regression(13, 32, output_size=1, num_layers=3).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=0.0001, betas=(0.9, 0.999), eps=1e-08, weight_decay=0) ds = TensorDataset(x_train,y_train) dl = DataLoader(ds,batch_size=32,shuffle=True) for i in range(2500): for j,(x,y) in enumerate(dl): x,y = x.to(device),y.to(device) out = model(x) loss = loss_function(out, y) loss.backward() optimizer.step() optimizer.zero_grad() train_loss.append(loss.item()) if j%10 == 0: print(f'epoch:{i+1} 第{j}次loss:{loss}') # 保存模型 torch.save(model.state_dict(),'lstm.pth') # torch.save(model.state_dict(),os.path.join(model_save_dir,model_file)) model.load_state_dict(torch.load('lstm.pth')) # 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 x_eval = x_eval.reshape(-1, 1, 13) # (seq_size, batch_size, feature_size) pred_test = model(x_eval) # 全量训练集 # 模型输出 (seq_size, batch_size, output_size) pred_test = pred_test.view(-1).cpu().data.numpy() print(x_eval.shape,pred_test.shape) # 反归一化 pred_test = inverse_transform_col(scaler,pred_test,-1) pred_test = pred_test.reshape(-1) y_eval = inverse_transform_col(scaler,y_eval,-1) # 打印指标 print(abs(pred_test - y_eval).mean() /y_eval.mean()) result_eight = pd.DataFrame({'pred_test': pred_test, 'real': y_eval}) target = (result_eight['pred_test'][-3:].sum() - result_eight['real'][-3:].sum()) / result_eight[ 'real'].sum() print(result_eight) print(target)