QRT Data Challenge: Liquid Asset Reconstruction

About this project

Participating in QRT's financial data competition to reconstruct missing liquid asset performance. Currently applying XGBoost and LightGBM with temporal features, volatility regimes, and anomaly detection logic. Exploring CNN + MLP hybrid models and feature selection to optimize RMSE. Ongoing experiment tracking and pipeline tuning in Python.

Key Features

• Time-Series Forecasting: Applies temporal machine learning to reconstruct missing asset returns.
• Feature Engineering: Builds volatility clusters, lag windows, macro regimes for predictive power.
• XGBoost & LightGBM: Runs boosting trees with cross-validation to minimize error.
• CNN + MLP Prototype: Experimental convolutional and dense neural net for signal refinement.