Weekly XAU/USD prediction API powered by stacked LSTM architecture with real-time macroeconomic integration. Built for institutional trading desks and algorithmic strategies.
Enterprise reliability meets cutting-edge deep learning. Every component optimized for institutional deployment.
Pure NumPy LSTM implementation (22KB weights). No TensorFlow runtime overhead. Instant response for high-frequency trading strategies.
Live Federal Reserve data (FFR) and Non-Farm Payrolls via FRED API. Automatically synced with weekly GC=F candles from Yahoo Finance.
2-layer LSTM network (24→12 units) captures short and medium-term dependencies. Ratio-based targets reduce non-stationarity sensitivity.
Validated on 10-week rolling backtest. MAE 3.7%, MAPE $162 average error. Consistent performance across volatile market conditions.
Flask + NumPy + scikit-learn. No GPU required. Deploys on 512MB RAM containers. Perfect for cost-effective cloud autoscaling.
Real-time performance monitoring with historical validation table. Track MAE, MAPE, RMSE, and direction accuracy across recent weeks.
End-to-end pipeline from market data ingestion to prediction API response.
Weekly OHLC candles (GC=F) from Yahoo Finance API with volume data.
FFR & NFP values from FRED aligned temporally using backward as-of merge.
9 ratio-based features (open/lag1, high/lag1, FFR change, returns) computed dynamically.
4-week sequence scaled, processed through stacked LSTM, denormalized to price projection.
Validated on live market data. Consistently outperforms naive baseline across multiple evaluation windows.
Integrate institutional-grade gold forecasting into your trading infrastructure. REST API and WebSocket streaming available.
