A preliminary analysis of deep learning and quantum machine learning approaches for stock prediction

Abstract

Predicting stocks has always been an attractive prospect for researchers as financial stock markets have been shown to have a significant economic impact on numerous sectors such as business, education, and technology. Techniques such as moving average models (e.g., autoregressive integrated moving average) and deep learning models (e.g., recurrent neural network) have been used to model financial markets. However, the stock market is incredibly volatile and there are no existing models that consistently model stocks. This study aims to reveal possible benefits of using quantum machine learning to model markets by proposing two models: a long short-term memory (LSTM) and a quantum neural network (QNN) consisting of two parametrized quantum circuits. The models were tested on an intraday dataset of Agilent Technologies, Inc., one of the index stocks on the S&P 500, with 1-minute intervals from 9-10-2017 to 2-12-2018. Each of these methods are evaluated using root mean squared error (RMSE) with the LSTM performing significantly better with an RMSE of 1.8695 (vs 4.4813 for QNN). Overall, quantum machine learning models still seem to be underdeveloped compared to classical models, and despite their faster training speeds, do not always perform better.