What Data Series Matter? Explaining key trends and factors generated by Artificial Intelligence

Abstract

We show a simple way to let the data speak for themselves. Specifically, we show how a large mixed bag of data, potentially embedded with missing data points and collinearities, and therefore unsuitable for traditional econometric analysis, can be useful in building fast and meaningful big data and artificial intelligence analyses and predictions. What’s more, our technique helps the results of the analyses to be easily interpreted by researchers. We use these techniques to build a surprisingly profitable E-mini crude oil futures trading strategy with monthly reallocations, delivering annualized returns of 100%+ with Sharpe ratio exceeding 2.2.