Pay Senedi Fiyat Yönünün Makine Öğrenmesi Yöntemleri ile Tahmini: Borsa İstanbul Örneği

Abstract

In this study, stock price with the calculated next three-month average of five manufacturing industry companies in the Borsa İstanbul 30 Index and the Corporate Governance Index was predicted with the data of the 2010/3 and 2020/3 periods. The dataset consisted of quarterly nine financial statements and five macroeconomic variables with a three-month average of the sample companies. Artificial Neural Networks, Classification and Regression Tree, and KNearest Neighbor Algorithm were used as prediction methods. A 10-fold cross-validation method was used in all methods in the study. In Artificial Neural Networks, Classification and Regression Tree analysis, the models that gave the best results in line with the given parameter ranges were obtained by using the determining the best parameters and performance criteria function. According to the results of the analysis, general classification accuracy was achieved 98.05% for Artificial Neural Networks, 96.10% for Classification and Regression Tree, and 92.20% K-Nearest Neighbor Algorithm. “Net Profit Margin”, “Price/Earning”, “Profit Per Share”, “CDS Premium (3-month average)”, “Consumer Confidence Index” were found as important variables that divided the data into two in the creation of the Classification and Regression Tree (CART) analysis. This result shows that the models used in this study can be incorporated into the models used by investors.