metadata
tags:
- patchtst
- transformer
- finance
- weekly-returns
- learnfinance
- time-series
LearnFinance PatchTST Model - v2026-01-30-a74b8b7aa0c3
Multi-signal PatchTST transformer model for predicting weekly stock returns.
Model Details
- Version: v2026-01-30-a74b8b7aa0c3
- Model Type: PatchTST (Patch Time Series Transformer)
- Training Window: 2011-01-01 to 2026-01-30
- Symbols: 15 stocks
Input Channels (11 total)
- OHLCV log returns (5): open, high, low, close, volume
- News sentiment (1)
- Fundamentals (5): gross_margin, operating_margin, net_margin, current_ratio, debt_to_equity
Metrics
- Train Loss: 0.009032704605066312
- Validation Loss: 0.016941080011589373
- Baseline Loss: 0.0006502264693262962
Usage
from brain_api.storage.huggingface import PatchTSTHuggingFaceModelStorage
storage = PatchTSTHuggingFaceModelStorage(repo_id="hajirazin/learnfinance-models-patchtst")
artifacts = storage.download_model(version="v2026-01-30-a74b8b7aa0c3")