Forecasting BIST 100 Index With Artificial Neural Networks and Regression Analysis

Abstract

Making reliable forecasts is very important for financial analysis. For this reason, financial analysts make analyzes using different models. Financial analysts try to make the most accurate estimation in these analyzes. The Artificial Neural Network model is a widely used method in the field of finance. In this study, BIST 100 index was estimated using Artificial Neural Networks and Regression model. By using the closing prices of the BIST 100 index between 2010 and 2020, the closing values of the BIST 100 index for 2021 were estimated. Moreover, the Regression model and Artificial Neural Network model predictions were obtained. The mean square error of the neural networks and Regression model was also found. Finally, according to the result of the mean of error squares, the performance of the models was compared and seen that the Artificial Neural Network model was better.

Improving Forecast Accuracy Using Combined Forecasts with Regard to Structural Breaks and ARCH Innovations

“…Telli and Coşkun (2016) also forecasted the BIST-100 Index using ANN models with daily data between July-November 2015 and showed the structured multilayer perceptron (MLP) model to be the best among the several tested models. Ünvan and Ergenç (2022) additionally compared the predictive ability of ANN and regression models applied to the BIST-100 Index's closing prices between 2010-2020 and found ANN to perform better than the considered regression models. Raşo and Demirci (2019) used deep learning methods to forecast the Turkish Stock Market on the BIST-30 Index from January 2016-April 2018, with their study's findings revealing the deep learning model to outperform other techniques such as support vector regression (SVR).…”