Beyond the Black Box: An Intuitive Approach to Investment Prediction with Machine Learning

The complexity of machine learning models presents a substantial barrier to their adoption for many investors. The algorithms that generate machine learning predictions are sometimes regarded as a black box and demand interpretation. In this article, the authors present a framework for demystifying the behavior of machine learning models. They decompose model predictions into linear, nonlinear, and interaction components and study a model’s predictive efficacy using the same components. Together, this forms a fingerprint to summarize key characteristics, similarities, and differences among different models. The presented framework is demonstrated in a case study applying random forest, gradient boosting machine, and neural network models to the challenge of predicting monthly currency returns. All models reliably identify intuitive effects in the currency market but also find new relationships attributable to nonlinearities and variable interactions. The authors argue that an understanding of these predictive components may help astute investors generate superior risk-adjusted returns. TOPICS:Statistical methods, simulations, big data/machine learning Key Findings • This article presents a framework for the implementation and interpretation of machine learning model predictions applied to investment portfolios. • Model predictions are decomposed into the linear, nonlinear, and interaction components, and their predictive efficacy is evaluated using these components. • Using a currency prediction case study, it is demonstrated that machine learning models reliably identify known effects and find new nonlinear relationships and interactions.