Comparing Decision Trees and Association Rules for Stock Market Expectations in BIST100 and BIST30

Abstract

With the increased financial fragility, methods have been needed to predict financial data effectively. In this study, two leading data mining technologies, classification analysis and association rule mining, are implemented for modeling potentially successful and risky stocks on the BIST 30 index and BIST 100 Index based on the key variables of index name, index value, and stock price. Classification and Regression Tree (CART) is used for classification, and Apriori is applied for association analysis. The study data set covered monthly closing values during 2013-2019. The Apriori algorithm also obtained almost all of the classification rules generated with the CART algorithm. Validated by two promising data mining techniques, proposed rules guide decision-makers in their investment decisions. By providing early warning signals of risky stocks, these rules can be used to minimize risk levels and protect decision-makers from making risky decisions.