Treffer: Auditor Choice Prediction using Machine Learning.

The purpose of this study was to predict the selection of independent auditors (BIG4 or non-BIG4) in the companies listed on the Warsaw Stock Exchange (WSE) using a Machine Learning approach. The population consisted of all the companies listed on the WSE during the period 1998-2021, except banks and insurance companies. The final sample included 11799 data specimens from 1074 companies during 23 consecutive years. The independent variables in the study were the financial ratios: log of revenue, log of assets, ROA, current ratio, and debt ratio and stock exchange (the main or alternative) of the sample companies. To analyze the data and predict the relevant class the following supervised learning techniques were employed: K-nearest neighbor classifier, Forest of trees, and Decision Tree. The models were built using the programming language Python and its libraries, including pandas, NumPy, matplotlib, and scikit-learn. The developed models had an accuracy of 85%. The results of this study showed that among the analyzed variables, company size measured by the log of assets and log of revenue is the most important feature to predict auditor choice. There are three main contributions of this study. First, it presents the determinants of auditor selection and extends the literature in building auditor choice models using machine learning algorithms. Second, it provides a methodology for the design and implementation ML approach for scientists. Third, it can be useful for audit practitioners, policymakers, investors, and other stakeholders to understand the factors affecting the selection of an independent auditor. The study is the first to focus on this topic in the specific context of Poland. [ABSTRACT FROM AUTHOR]