Predicting S&P 500 Index ETF (SPY) During COVID-19 via K-Nearest Neighbors (KNN) Algorithm

Abstract

In this paper, the daily adjusted closing price of SPY (SPDR S&P 500 ETF Trust) is predicted by using the High-Low prices of SPY, DIA (SPDR Dow Jones Industrial Average ETF Trust), and QQQ (Invesco NASDAQ-100 ETF Trust) via the KNN method during the COVID-19 pandemic period. Results show that applying the KNN method, a simple, intuitive, and explainable machine learning method, is feasible and effective in SPY price prediction and corresponding trade decisions during the COVID-19 pandemic. Experiments also indicate that adding information on High-Low prices from DIA (a value tilt ETF) and QQQ (a growth tilt ETF) cannot improve the accuracy of both SPY price prediction and trading decisions. Results are consistent with previous findings based on the portfolio approach that value spread does not help predict stock market returns.