Performance comparison of classification techniques in stock exchange index direction movement prediction: the case of BIST 100

The aim of this study is to show that the direction of movement of stock market index is predictable by using technical indicators as input data and to compare performances of classification methods. The BIST 100 index direction was estimated by using the daily closing data of Borsa Istanbul 100 Index (BIST 100) of 10 technical indicators belong to the period of 1995:3-2018:3. By using the Machine learning methods such as Artificial Neural Networks (ANN), Support Vector Machines (SVM), Decision Trees (DT), Naive Bayes (NB), k-Nearest Neighborhood (k-NN) methods and the statistical methods of Logistic Regression (LogR) and Linear Discriminant Analysis (LDA) were analyzed. The correct classification rates of classification methods were 83.83%, 78.43%, 65.04%, 61.74%, 55.48%, 76.70% and 76.87% respectively. According to the results, ANN is the best classification method that can be used to predict the direction of movement of the BIST 100 index.