Predicting VIX with Adaptive Machine Learning

Abstract

We study economic predictors of the CBOE implied volatility index (VIX). Designing an automated machine learning framework enables us to study a comprehensive list of 278 variables for the first time. Adaptive Boosting emerges as the best classification model chosen at the validation stage which also has good performance during the 11-year out-of-sample period. Our tests on the source of predictability show both the nonlinear methods and the comprehensive economic variables are important. Although VIX spikes are not predictable, the algorithms can adapt quickly by abstracting new information from the data to recover from the losses in trading strategy tests. Besides the modelling techniques, the weekly US jobless report is the most important source of predictability along with some S&P 500 members’ technical indicators.

Video abstract: https://youtu.be/L_FQ6dFuYr0

Keywords: Machine Learning, AutoML, Explainable AI, VIX, Predictability, Forecasting, Quantitative Trading, Big Data, S&P 500, Futures, US markets

JEL Classification: G0, G17, C52, C55, C58