🚀 BTCUSDT AI Ensemble — LSTM + GRU + Dense (TF.js) — Single File
Disclaimer: This is a research/demo system. It fetches public Binance market data for analysis/backtests only.
It is not financial advice. No system can guarantee 95% winning trades — results vary by market, latency, fees, and risk settings.
Idle
This demo trains small models in-browser (TF.js). For real production, train offline on GPU/TPU and only use vetted weights in live systems.
Model Hyperparams & Tuning
Randomized search across neurons & dropout per trial. Early stopping implemented. Models: LSTM, GRU, Dense proxy (XGBoost-like).
Backtest Results & Controls
No runs yet
| # | Time | Side | Entry | Pred | Profit | Conf |
|---|
Logs
Ensemble Weights & RL
Weights initialized equal; after each resolved simulated trade a lightweight multiplicative reward raises weight for models that predicted correctly.
Current weights
LSTM: 33% • GRU: 33% • DENSE: 34%
Comments
Post a Comment