Classification Of BIST -100 Index’ Changes Via Machine Learning Methods

Abstract

The changes in BIST-100 index are economically crucial. In this study, classifications will be made with the assumption that the changes in BIST-100 index are dependent on certain factors. The classifiers to be used are k-nearest neighbor algorithm, naive Bayes Classifier, logistic regression and C4.5 classifier from the machine learning methods. Factors affecting the change of BIST-100 index values are deemed as Euro/ Dollar Parity, Gold value (ounce), Crude Oil Prices, Monthly Interest Rates, Inflation Data and DAX, FTSE, S&P 500 that are widely used in the literature. As a result of the transactions performed via Weka program, the most successful methods in order are C4.5 classifier algorithm (66.2%) and logistic regression analysis (65.9%).