Auditor's Opinions Prediction with Machine Learning Algorithms

Abstract

The goal of this study is to create a model to predict auditor opinions and to check the accuracy of previous ones. The data for the years 2010-2019 have been collected from the Public Disclosure Platform (KAP) of Turkey, and financial ratios for relevant periods have been calculated from financial statements. The collected and calculated data have been combined with other additional information about financial report progress, and a data set consist of 7128 rows and 52 feature has been obtained. The data set is the largest data set in the literature regarding publicly-traded companies. It has been used to classify auditor opinions. As a result of the studies, when compared other tested algorithms, the XGBoost algorithm has given the highest Fl-score with %94.7. The proposed model in this study is an auxiliary tool for independent auditors for risk assessment and quality control studies.