Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Advanced Quantitative Factor Research: ML-powered stock return prediction with 72% performance improvement. Features comprehensive alpha factor library, systematic feature selection, and deep learning models (LSTM+ResNet achieving IC=0.06476).
| Date | Stars |
|---|---|
| 2026-07-24 | 413 |
| 2026-07-25 | 413 |
| 2026-07-28 | 413 |
| 2026-07-30 | 413 |
| 2026-07-31 | 415 |
| 2026-08-06 | 415 |
Today
— stars today
This week
+2 stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.48%/day
# Quantitative Factor Research Project A comprehensive quantitative finance research project implementing **factor modeling**, **feature selection**, and **machine learning** for stock return prediction. This project demonstrates advanced techniques in alpha factor generation, systematic feature selection, and deep learning model comparison. ## 🎯 Project Overview This project implements a complete quantitative research pipeline covering the entire workflow from raw market data to production-ready predictive models. The research achieved **significant performance improvements** with the best model (LSTM+ResNet) reaching an **Information Coefficient of 0.06476**, representing a **72% improvement** over traditional linear regression benchmarks. ## 📊 Key Achievements - **🏗️ Factor Library**: Generated comprehensive alpha factor library using vectorized operations and technical operators - **🔍 Feature Selection**: Systematically reduced features from 100+ candidates to 85 high-quality factors using statistical and ML methods - **🤖 Model Performance**: Achieved IC of 0.06476 with LSTM+ResNet architecture, significantly outperforming baseline models - **⚡ Optimization**: Implemented automated hyperparameter tuning using Optuna for all model architectures ## 🏗️ Project Structure ``` quantitative-factor-research/ ├── notebooks/ # Jupyter notebooks for analysis │ ├── Data_Preparation.ipynb # Data loading and preprocessing │ ├── Alpha_Factor_Generation.ipynb # Base alpha factor creation │ ├── Alpha_Factor_Generation_2.ipynb # High-frequency technical factors │ ├── Alpha_Factor_Selection.ipynb # Feature selection pipeline │ ├── ML_Model.ipynb # Machine learning models │ └── factor_backtest.ipynb # Factor backtesting framework ├── data/ # Data storage │ ├── raw/ # Raw market data │ ├── processed/ # Processed datasets │ ├── factors/ # Generated factors │ │ ├── obtained_features/ # Raw factor library │ │ └── selected_factors/ # Final selected factors │ └── backtest_results/ # Backtesting outputs └── README.md # Project documentation ``` ## 🔬 Research Methodology ### 1. Factor Modeling **Objective**: Develop a comprehensive factor library capturing various market dynamics and price-volume relationships. **Implementation**: - **Technical Operators**: Implemented time-series operators (ts_max, ts_min, ts_sum, ts_std_dev, ts_delta, delay, rank) - **Vectorized Operations**: Utilized pandas and numpy for efficient large-scale factor computation - **Wide-table Dataset**: Generated standardized factor matrices aligned across time and assets - **Base Alpha Enhancement**: Applied quantile-based segmentation to improve factor discriminative power - **Dynamic Factors**: Incorporated rolling standard deviation and price-volume correlation factors **Key Features Generated**: - **Classic Alpha101 Factors**: Traditional quantitative factors - **Technical Indicators**: RSI, MACD, Bollinger Bands variations - **Volume Microstructure**: Intraday volume patterns and clustering - **Price-Volume Relationships**: Correlation-based and interaction factors - **Statistical Moments**: Skewness, kurtosis, and higher-order moments ### 2. Feature Selection **Objective**: Systematically reduce the feature space while maintaining predictive power and ensuring factor independence. **Multi-Stage Selection Process**: 1. **Statistical Screening**: - **Information Coefficient (IC)**: Filtered factors with |IC| > 0.02 - **IC Consistency**: Maintained factors with stable predictive relationships 2. **Correlation Analysis**: - **Spearman Correlation**: Eliminated highly correlated factors (threshold < 0.7) - **Greedy Selection**: Iterative removal to maintain factor diversity 3. **Machine Learning Selection**:
Excerpt of 11,096 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:c54aed2d95af98cc, topic:deep-learning, topic:pytorch
matched fp:c54aed2d95af98cc, topic:transformer