ARTIFICIAL NEURAL NETWORK (ANN): PORTFOLIO OPTIMIZATION AND TRACKING ERROR

Abstract

This study aims to investigate the application of the Artificial Neural Network (ANN) model in portfolio optimization, consider tracking errors in portfolio optimization, and evaluate their performance. The methodology comprises of two stages. At first, developing optimized portfolios based on market capitalization and risk using the ANN model. Second, optimizing portfolios while considering tracking error with the ANN model. It utilizes monthly prices data of listed companies on the Pakistan Stock Exchange and the KSE-100 index as a benchmark, ranging from December 31^st, 1999, to December 31^st, 2023. Performance Metrics like Mean Squared Error (MSE), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE) assess prediction accuracy. The findings of this study indicate consistently low Mean Squared Error (MSE) across all portfolios, affirming the efficacy of the ANN model in portfolio optimization and considering tracking error in portfolio optimization. Comparing the tracking error portfolios with market capitalization and risk portfolios reveals that the MSE values for tracking error portfolios are lower. This suggests that tracking error portfolios perform better. Researchers and financial analysts used traditional approaches to build portfolios to balance risk and return that depend on fixed assumptions. This research adds value to the existing literature on performance measures. Furthermore, this research uses the ANN model in portfolio optimization while considering tracking errors.