Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
AI Research
arXiv CS.AI·28d ago·5 min read
A Hybrid LSTM-XGBoost Framework for Multi-Horizon Stock Return Prediction Across Diversified Equity Portfolios
TL;DR
A hybrid framework combining LSTM and XGBoost was developed for predicting stock returns across 14 U.S. equities. It achieved a test RMSE of 0.0949 for 30-day predictions, significantly outperforming standalone models.
✦ Why It Matters
Implement this hybrid model to enhance your stock return predictions and improve investment strategies today.
Key Takeaways
- Utilize LSTM for sequential market feature processing.
- Combine LSTM embeddings with technical indicators for better predictions.
- Optimize XGBoost using cross-validation for improved accuracy.
- Apply the investment scoring framework for portfolio decision-making.
How It Works
The hybrid framework first employs an LSTM network to analyze 60-day sliding windows of market data, generating embeddings that capture temporal dynamics. These embeddings are then concatenated with hand-crafted technical indicators, creating a rich feature set for the XGBoost regressor, which is trained to predict stock returns over multiple horizons.