S&P/BMV IPC Forecasting Using Quantum Long Short-Term Memory

Abstract

This paper presents results on time seriesforecasting using a Quantum Long Short-Term Memory,known by its acronym QLSTM. We present experimentalresults about the forecasting on the Mexican financialclosing price dataset (S&P/BMV IPC), In addition toalso making a regression on the time series related tothe closing value of companies relevant to this financialindex, where their probability distributions and theircorrelation dimensions are analyzed to validate that theyshow behaviors that are not linear and persistent, whichdescribe their complex behavior and reinforce the ideaof using more appropriate tools such as quantum neuralnetworks.To evaluate the performance of the QLSTM, we makea continuous comparison of the results obtained by itsclassical counterpart, the LSTM neural network. It isnoteworthy that this work represents the first publicationin non-linear time series prediction applied to a Mexicanstock index using quantum computing.