Deep Learning-based Decision Integration Strategy for the Price Prediction of Istanbul Stock Exchange (BIST100)

To predict the movement direction and price of stock markets, exchange rates, stocks, and commodities are important research field for researchers, analysts, and investors. In this research, a deep learning-based decision consolidation approach is proposed for forecasting the price of Istanbul Stock Exchange (BIST 100). For this purpose, the dataset is gathered between February 2011 and February 2021 by including various features. In addition to the stock market index information of countries in Europe, Asia and the USA, additional features that we consider have impact on the direction of trend of the Istanbul Stock Exchange have been added to the data set. In order to estimate the price of BIST100, multilayer perceptron (MLP), convolutional neural network (CNN), and long short-term memory network (LSTM) are evaluated as deep learning-based strategy. For the purpose of obtaining better prediction performance, decision integration strategy is assessed by employing gradient boosting regressor approach. Experiment results indicate the combination of deep learning models exhibits a substantial potential to forecast the price of BIST 100.