XAUUSD H1 LSTM Classifier with Trading Breakout Strategies

• Dataset: OHLCV data with ~200k rows
• Input: 60-timestep sequences with Open, High, Low, Close, Volume, and breakout features
• Breakout Strategies:
  • Breakout Pullback: Detects pullbacks after breakouts
  • Pattern #1: Key support/resistance with minimum two touches
  • Pattern #2: Reduction in pullback reactions
• Task: Classification (Buy: >0.1% increase, Sell: >0.1% decrease, Hold: otherwise)
• Training: 10 epochs, Adam optimizer, CrossEntropyLoss with class weights
• Best Test Accuracy: [insert best_accuracy]
• Author: Kasemsan (Ken)
• GitHub: semkken
• Homepage: spacevalu.com
• Usage:
  python

  1import torch, joblib 2class ForexLSTMClassifier(torch.nn.Module): 3 def __init__(self, input_size=13, hidden_size=64, num_layers=2, num_classes=3): 4 super(ForexLSTMClassifier, self).__init__() 5 self.lstm = torch.nn.LSTM(input_size, hidden_size, num_layers, batch_first=True, dropout=0.2) 6 self.fc = torch.nn.Linear(hidden_size, num_classes) 7 self.softmax = torch.nn.Softmax(dim=1) 8 def forward(self, x): 9 out, _ = self.lstm(x) 10 out = self.fc(out[:, -1, :]) 11 out = self.softmax(out) 12 return out 13model = ForexLSTMClassifier() 14model.load_state_dict(torch.load("pytorch_model.bin")) 15scaler = joblib.load("scaler.pkl") 16# Prepare 60-timestep input and predict