Market Regime Prediction

ML Engineer

Developed an end to end market regime prediction framework using machine learning models. We started with a dataset of 100M+ time-series datapoints across 10+ years of stock market data. This was cleaned, processed, and feature engineered to be used in machine learning models. We used several models such as GMM, Hierarchical clustering, LSTM, and logistic regression. The model performance was evaluated using measurements such as silhouette score, Davies-Bouldin index, confusion matricies, and directional accuracy resulting in 58-62% prediction accuracy.

Conducted as a group project for Machine Learning (CS4641) at The Georgia Institute of Technology