Add comprehensive README with model details, metrics, and usage instructions
Browse files
README.md
ADDED
|
@@ -0,0 +1,96 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
language: en
|
| 4 |
+
library_name: stable-baselines3
|
| 5 |
+
tags:
|
| 6 |
+
- reinforcement-learning
|
| 7 |
+
- finance
|
| 8 |
+
- gold-trading
|
| 9 |
+
- xauusd
|
| 10 |
+
- ppo
|
| 11 |
+
metrics:
|
| 12 |
+
- sharpe_ratio
|
| 13 |
+
- win_rate
|
| 14 |
+
pipeline_tag: reinforcement-learning
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
# PPO Model for XAUUSD Gold Trading
|
| 18 |
+
|
| 19 |
+
This repository contains a Reinforcement Learning model trained using Proximal Policy Optimization (PPO) for trading XAUUSD (Gold vs US Dollar) on 15-minute timeframes.
|
| 20 |
+
|
| 21 |
+
## Model Details
|
| 22 |
+
|
| 23 |
+
- **Model Type**: PPO (Proximal Policy Optimization)
|
| 24 |
+
- **Framework**: Stable-Baselines3
|
| 25 |
+
- **Environment**: Custom Gym environment for XAUUSD trading
|
| 26 |
+
- **Training Data**: Historical XAUUSD data from 2004 to 2025 (resampled to 15-min bars)
|
| 27 |
+
- **Total Timesteps**: 1,000,000
|
| 28 |
+
- **Position Sizing**: Base 5.0 oz, Max 7.5 oz
|
| 29 |
+
- **Initial Capital**: 200 USD
|
| 30 |
+
- **Transaction Cost**: 0.65 USD per oz
|
| 31 |
+
|
| 32 |
+
## Performance Metrics (Test Set)
|
| 33 |
+
|
| 34 |
+
- **Average Daily Profit**: 51.46 USD
|
| 35 |
+
- **Win Rate**: 69.0%
|
| 36 |
+
- **Max Drawdown**: 12.0%
|
| 37 |
+
- **Sharpe Ratio**: 7.56
|
| 38 |
+
- **Average Trades per Day**: 2.66
|
| 39 |
+
|
| 40 |
+
## Features Used
|
| 41 |
+
|
| 42 |
+
- Log Return
|
| 43 |
+
- RSI (14-period)
|
| 44 |
+
- Moving Averages (short/long)
|
| 45 |
+
- Bollinger Bands
|
| 46 |
+
- MACD
|
| 47 |
+
- Volume indicators
|
| 48 |
+
|
| 49 |
+
## Usage
|
| 50 |
+
|
| 51 |
+
### Loading the Model
|
| 52 |
+
|
| 53 |
+
```python
|
| 54 |
+
from safetensors.torch import load_file
|
| 55 |
+
from stable_baselines3 import PPO
|
| 56 |
+
import torch
|
| 57 |
+
|
| 58 |
+
# Load state dict from safetensors
|
| 59 |
+
state_dict = load_file("ppo_xauusd.safetensors")
|
| 60 |
+
policy = PPO.policy_class(observation_space, action_space) # Define spaces accordingly
|
| 61 |
+
policy.load_state_dict(state_dict)
|
| 62 |
+
|
| 63 |
+
# Create model
|
| 64 |
+
model = PPO(policy=policy, env=env) # Or load full model if available
|
| 65 |
+
```
|
| 66 |
+
|
| 67 |
+
### For Full Inference
|
| 68 |
+
|
| 69 |
+
To use the model for trading, you'll need to:
|
| 70 |
+
1. Set up the trading environment (`XAUUSDTradingEnv`)
|
| 71 |
+
2. Load VecNormalize stats
|
| 72 |
+
3. Run predictions
|
| 73 |
+
|
| 74 |
+
Note: This is a simulation model. Use with caution in real trading.
|
| 75 |
+
|
| 76 |
+
## Training Configuration
|
| 77 |
+
|
| 78 |
+
- Learning Rate: 0.0003
|
| 79 |
+
- Batch Size: 256
|
| 80 |
+
- Gamma: 0.99
|
| 81 |
+
- GAE Lambda: 0.95
|
| 82 |
+
- Clip Range: 0.2
|
| 83 |
+
- Entropy Coefficient: 0.01
|
| 84 |
+
|
| 85 |
+
## Files
|
| 86 |
+
|
| 87 |
+
- `ppo_xauusd.safetensors`: Model weights in SafeTensors format
|
| 88 |
+
- `vecnormalize.pkl`: VecNormalize statistics for observation normalization
|
| 89 |
+
|
| 90 |
+
## License
|
| 91 |
+
|
| 92 |
+
MIT License
|
| 93 |
+
|
| 94 |
+
## Disclaimer
|
| 95 |
+
|
| 96 |
+
This model is for educational and research purposes only. Trading involves risk, and past performance does not guarantee future results. Always backtest and validate before using in live trading.
|