Integration of modified firefly algorithm and machine learning regression for stock price forecasting

Abstract

Stock price prediction plays an important financial topic, and is received attention from researchers and practitioners. Therefore, creating an intelligent model that can accurately forecast stock price in a robust way has always been a great interest for investors and financial analysts. The objective of this paper is to propose a developed sliding-window machine learning regression model for forecasting stock prices. The proposed model integrates the Lévy flight (LF), firefly algorithm (FA) and the least squares support vector regression (LSSVR). The LF integrated into the FA automatically fine-tunes the hyperparameters of the LSSVR to construct an enhanced LSSVR model. The developed model is called the LFFA-LSSVM. This investigation uses stock popular datasets; namely ISE100, BIST 100 and NASDAQ; to verify the effectiveness of the proposed model. The results revealed that for ISE100 stock price dataset the LFFA-LSSVR model obtained highest performance with a root mean square error (RMSE) of 0.108, a mean absolute error (MAE) of 0.079, an mean absolute percentage error (MAPE) of 0.578%, and an correlation coefficient (R) of 0.983. Therefore, the proposed model is considered as a suitable tool for predicting stock prices.