← Back

Project

Generating Alpha with Multihorizon Forecasting

Solo project · Dec 2024 - Mar 2025 · Quant research

Analyzed NIFTY index returns using macroeconomic and inter-market data, the S&P 500, India VIX, USD/INR volatility, bond yields, and gold ratios, across weekly and monthly horizons.

How it was validated

  1. 01

    Multi-source features. Macro and inter-market signals, the S&P 500, India VIX, USD/INR volatility, bond yields and gold ratios, tested across both weekly and monthly horizons.

  2. 02

    Full residual diagnostics. Ran Shapiro-Wilk, Breusch-Pagan and Durbin-Watson tests to validate regression assumptions before trusting any of the output.

  3. 03

    Ridge over Lasso. Compared both, and chose Ridge for its stability in a multi-factor setting, Lasso's penalization was too aggressive and zeroed out useful signal.

Ridge Regression Lasso Regression Time Series Finance

A note on scope: the alpha-generation and deployment part of this project is temporarily taken down, it wasn't performing reliably and I'd rather pull it than leave something broken live. The forecasting and diagnostics work above stands on its own.

View on GitHub