Hands-On Machine Learning for Algorithmic Trading

Feature engineering and alpha factor research

More than the raw data, feature engineering is often the key to making signal useful for an algorithm. Leveraging decades of research into risk factors that drive returns on theoretical and empirical grounds is a good starting point to prioritize data transformations that are more likely to reflect relevant information.

However, only creative feature engineering will lead to innovative strategies that can compete in the market over time. Even for new alpha factors, a compelling narrative that explains how they work, given established ideas on market dynamics and investor behavior, will provide more confidence to allocate capital.

The risks of false discoveries and overfitting to historical data ...