semsankken/xauusd_forex_lstm

XAUUSD-H1 Forex LSTM Model

Fine-tuned LSTM for predicting next hourly Close price on XAUUSD (gold) data from 2008-2025.

• Dataset: Hourly OHLCV from "XAUUSD_H1.csv" (~150k rows)
• Input: Sequences of 60 timesteps with Open, High, Low, Close, Volume, SMA_20
• Task: Regression (next Close price)
• Training: 100 epochs, Adam optimizer, MSE loss
• Evaluation: Test MSE ~ 0.9100
• Usage: Load with PyTorch, use scaler for inverse transform.

Inference Example

import torch
import json
import joblib
import numpy as np

class ForexLSTM(torch.nn.Module):
    # ... (copy class definition)

model = ForexLSTM()
model.load_state_dict(torch.load("pytorch_model.bin"))
scaler = joblib.load("scaler.pkl")
# Prepare input sequence and predict as above