STOCK PRICE PREDICTION USING OLL NEURAL NETWORK COMPARED WITH MULTIPLE LINEAR REGRESSION

Abstract

Predicting the direction of stock price movement has been one of the most widely investigated and challenges for both investors and scholars.A considerable body of research has been devoted to the analysis of stock markets, leveraging advanced methodologies hailing from mathematical, computational, economic, and diverse interdisciplinary domains.This study introduces the Optimization Layer by Layer Neural Network (OLLNN) as a modern approach aimed at refining the weight adjustment process within each layer through the resolution of linear problems, thereby enhancing modeling precision and expediting convergence.To empirically evaluate the OLLNN's efficacy, we conducted a case study of PTTXP firm, a globally significant oil and gas production company with a multi-billion-dollar valuation, employing it as a subject for predicting daily closing stock prices.The predictive variables encompass a group of energy commodity daily prices such as Brent oil, WTI oil, and LNG prices, in conjunction with PTTXP's own stock price, some stock index, and associated indicators.Furthermore, we employ Multiple Linear Regression (MLR) as a benchmark for prediction accuracy.A comparative analysis of the results indicates that both methods yield similar Mean Absolute Percentage Error (MAPE) values, approximating around 1.00%.Moreover in validation, the OLLNN performs prediction of new dataset with the highest accuracy of 0.87% error.These findings underscore the potential of OLLNN as a valuable tool for enhancing the accuracy of stock price movement predictions.