Prediction of Independent Audit Firm Switching By Using Machine Learning Methods: The Case of Türkiye

This study aimed to predict independent audit firm switching of the companies traded in Borsa Istanbul Star Market (BIST STARS) in Türkiye by using financial ratios and machine learning algorithms. In this context, 13 financial datasets of 158 companies traded in BIST STARS in the 2019-2021 period were used as input variables. First, the significance values of the input variables were found by using the Mutual Information (MI) method. Then, input variables were grouped sequentially in order of importance to select the most accurate subset representing the data. Among the machine learning algorithms, Support Vector Machine, Decision Tree, Random Forest, Naive Bayes,K-Nearest Neighbors, and XGBoost algorithm methods were used for group selection. GridSearchCV technique was applied to optimize the initial parameters of the methods. As a result of the experiments, the XGBoost algorithm was found to be the most successful method in predicting the change of independent audit firm with an accuracy value of 88.4%. It was sufficient for the method to use 8 attributes selected from 13 financial datasets. On the other hand, the Return on Assets (ROA) was determined as the most important attribute.