skka3134

skka3134

email
telegram

Machine Learning and Quantitative Investing: 4. Building LSTM

  1. Choose to use CPU or GPU for training
if torch.cuda.is_available():
	device = torch.device("cuda")
else:
	device = torch.device("cpu")
  1. Build LSTM model
class LSTM(nn.Module):
    def __init__(self, input_size, hidden_size, num_stacked_layers,output_size):
        super().__init__() # Initialize the constructor of the parent class
        self.hidden_size = hidden_size
        self.num_stacked_layers = num_stacked_layers
        self.lstm = nn.LSTM(input_size, hidden_size, num_stacked_layers, batch_first=True) # Construct the LSTM model
        self.fc = nn.Linear(hidden_size,output_size) # Fully connected layer
    # Forward propagation
    def forward(self, x):
        batch_size = x.size(0)
        # Initialize the hidden state
        h0 = torch.zeros(self.num_stacked_layers, batch_size, self.hidden_size).to(device)
        c0 = torch.zeros(self.num_stacked_layers, batch_size, self.hidden_size).to(device)
        out, _ = self.lstm(x, (h0, c0)) # Separate the hidden state to avoid gradient explosion
        out = self.fc(out[:, -1, :]) # Only take the last hidden state
        return out 

# Initialize the LSTM model
input_size=1 # Input dimension, close
hidden_size=4 # Hidden layer dimension
num_stacked_layers=1  # Number of LSTM layers
output_size=1 # Output dimension, close
model = LSTM(input_size,hidden_size,num_stacked_layers,output_size)
# model.to(device)

Parameter settings

# Define the learning rate
learning_rate = 0.001
# Define the loss function
loss_function = nn.MSELoss() # nn.CrossEntropyLoss is commonly used for binary classification problems, nn.NLLLoss is commonly used for image recognition
# Define the optimizer
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate) # Choose the optimization algorithm, see https://blog.csdn.net/S20144144/article/details/103417502 for more information
  1. Training
def train_one_epoch():
    model.train(True)
    print(f'Epoch: {epoch + 1}')
    running_loss = 0.0
    
    for batch_index, batch in enumerate(train_loader):
        x_batch, y_batch = batch[0].to(device), batch[1].to(device)

        output = model(x_batch) # Forward propagation
        loss = loss_function(output, y_batch) # Calculate the loss
        running_loss += loss.item()
        optimizer.zero_grad() # Gradient accumulation, clear the gradient
        loss.backward() # Backward propagation
        optimizer.step() # Update parameters

        if batch_index % 100 == 99:  # print every 100 batches
            avg_loss_across_batches = running_loss / 100
            print('Batch {0}, Loss: {1:.3f}'.format(batch_index+1,avg_loss_across_batches))
            running_loss = 0.0
    print()
  1. Validation
def validate_one_epoch():
    model.train(False)
    running_loss = 0.0
    # Iterate through the test set, get data, and make predictions
    for batch_index, batch in enumerate(test_loader):
        x_batch, y_batch = batch[0].to(device), batch[1].to(device)
        with torch.no_grad():
            output = model(x_batch)
            loss = loss_function(output, y_batch)
            running_loss += loss.item()
    # Calculate
    avg_loss_across_batches = running_loss / len(test_loader)
    # Print
    print('Val Loss: {0:.3f}'.format(avg_loss_across_batches))
  1. Define the number of training epochs
num_epochs = 10
for epoch in range(num_epochs):
    train_one_epoch()
    validate_one_epoch()
with torch.no_grad():
    predicted = model(X_train.to(device)).to('cpu').numpy()

image
5. Visualization

plt.plot(y_train, label='Actual Close')
plt.plot(predicted, label='Predicted Close')
plt.xlabel('Day')
plt.ylabel('Close')
plt.legend()
plt.show()

image

Loading...
Ownership of this post data is guaranteed by blockchain and smart contracts to the creator alone.