输出预测结果

浙江行业电量/分类归一化.py

@@ -2,6 +2,27 @@ import os
 import numpy as np
 import pandas as pd
 from sklearn.preprocessing import MinMaxScaler
+import torch
+from torch import nn
+from torch.utils.data import DataLoader, TensorDataset
+import matplotlib.pyplot as plt
+
+class LSTM(nn.Module):
+    def __init__(self, input_size, hidden_size, output_size, num_layers):
+        super().__init__()
+        self.lstm = nn.LSTM(input_size, hidden_size, num_layers)
+        self.fc1 = nn.Linear(hidden_size, 128)
+        self.fc2 = nn.Linear(128, output_size)
+        self.ReLu = nn.ReLU()
+        self.dropout = nn.Dropout(0.5)
+
+    def forward(self, x):
+        x, _ = self.lstm(x)
+        s, b, h = x.shape
+        x = x.reshape(-1, h)
+        output = self.ReLU(self.dropout(self.fc1(x)))
+        output = self.fc2(output)
+        return output
 
 
 def normal(data):
@@ -63,7 +84,7 @@ dataset_x, dataset_y = create_dataset(df_city_industry)
 for city in df['地市'].drop_duplicates()[1:]:
     df_city_industry = df[df['地市'] == city][industry]
     x, y = create_dataset(df_city_industry)
-    dataset_x,dataset_y = np.concatenate([dataset_x,x]),np.concatenate([dataset_y,y])
+    dataset_x, dataset_y = np.concatenate([dataset_x, x]), np.concatenate([dataset_y, y])
 
 for industry in df.columns[2:][1:]:
     for city in df['地市'].drop_duplicates():
@@ -73,4 +94,61 @@ for industry in df.columns[2:][1:]:
 
 print(dataset_x.shape, dataset_y.shape)
 
-
+train_size = int(0.7 * len(dataset_x))
+x_train, y_train = dataset_x[:train_size], dataset_y[:train_size]
+x_eval, y_eval = dataset_x[train_size:], dataset_y[train_size:]
+x_train, y_train = torch.from_numpy(x_train).type(torch.float32), torch.from_numpy(y_train).type(torch.float32)
+x_eval, y_eval = torch.from_numpy(x_eval).type(torch.float32), torch.from_numpy(y_eval).type(torch.float32)
+
+ds = TensorDataset(x_train, y_train)
+dl = DataLoader(ds, batch_size=128, shuffle=True, drop_last=True)
+eval_ds = TensorDataset(x_eval, y_eval)
+eval_dl = DataLoader(eval_ds, batch_size=256, drop_last=True)
+
+device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+model = LSTM(10,64, 3, num_layers=2).to(device)
+loss_fn = nn.MSELoss()
+optimizer = torch.optim.Adam(model.parameters(), lr=0.0001)
+
+min_loss = 1
+for epoch in range(10):
+    for step, (x, y) in enumerate(dl):
+        x, y = x.to(device), y.to(device)
+        pred = model(x)
+        loss = loss_fn(pred,y)
+        optimizer.zero_grad()
+        loss.backward()
+        optimizer.step()
+
+        if (step+1) % 1000 == 0:
+            print(f'epoch{epoch+1}: train_step{step}/{len(dl)} train_loss:{round(loss)}\n')
+
+    model.eval()
+    batch_loss = 0
+    with torch.no_grad():
+        for x,y in eval_dl:
+            x, y = x.to(device), y.to(device)
+            pred = model(x)
+            loss = loss_fn(pred, y)
+            batch_loss += loss
+    print(f'epoch{epoch+1}: eval_loss:{batch_loss/len(eval_dl)}')
+
+    if batch_loss/len(eval_dl) < min_loss:
+        min_loss = batch_loss/len(eval_dl)
+        best_parameters = model.state_dict()
+
+torch.save(best_parameters,'best_3.pth')
+
+model = LSTM(10,64, 3, num_layers=2).to(device)
+model.load_state_dict(torch.load('best_3.pth'))
+dataset_x = dataset_x.reshape(-1,1,10)
+dataset_x = torch.from_numpy(dataset_x).type(torch.float32).to(device)
+pred = model(dataset_x).reshape(-1)
+pred = np.concatenate((np.zeros(10), pred.cpu().detach().numpy()))
+
+
+plt.plot(pred, 'r', label='prediction')
+plt.plot(dataset_y.reshape(-1), 'b', label='real')
+plt.plot((train_size*3, train_size*3), (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()