|
|
--- |
|
|
license: mit |
|
|
tags: |
|
|
- finance |
|
|
- stock-prediction |
|
|
- forex |
|
|
- time-series |
|
|
- pytorch |
|
|
- ensemble-learning |
|
|
library_name: pytorch |
|
|
--- |
|
|
|
|
|
# ARA AI - Financial Prediction Models |
|
|
|
|
|
Continuously trained ensemble ML models for stock and forex prediction. |
|
|
|
|
|
## Models |
|
|
|
|
|
- **Stock Models**: Trained on random selection of stocks every 2 hours |
|
|
- **Forex Models**: EURUSD, GBPUSD, USDJPY trained every 2 hours |
|
|
|
|
|
## Training Schedule |
|
|
|
|
|
Models are automatically retrained every 2 hours (12 times daily) using: |
|
|
- 2 years of historical data |
|
|
- Incremental training on existing models |
|
|
- Ensemble of XGBoost, LightGBM, Random Forest, Transformers, CNN-LSTM |
|
|
|
|
|
## Usage |
|
|
|
|
|
```python |
|
|
from huggingface_hub import hf_hub_download |
|
|
from meridianalgo.unified_ml import UnifiedStockML |
|
|
|
|
|
# Download model |
|
|
model_path = hf_hub_download( |
|
|
repo_id="MeridianAlgo/ARA.AI", |
|
|
filename="models/stock_AAPL.pt" |
|
|
) |
|
|
|
|
|
# Load and predict |
|
|
ml = UnifiedStockML(model_path=model_path) |
|
|
prediction = ml.predict('AAPL', days=5) |
|
|
``` |
|
|
|
|
|
## Experiment Tracking |
|
|
|
|
|
Training metrics tracked on [Weights & Biases](https://wandb.ai) |
|
|
|
|
|
## Repository |
|
|
|
|
|
Source code: [github.com/MeridianAlgo/AraAI](https://github.com/MeridianAlgo/AraAI) |
|
|
|
|
|
## Disclaimer |
|
|
|
|
|
These models are for educational and research purposes only. Not financial advice. |
|
|
Past performance does not guarantee future results. |
|
|
|
