# Forecasting auditor’s going concern opinion using with hybrid robust machine learning model

## TL;DR

This paper proposes a hybrid machine learning model to accurately predict auditors' going concern opinions using data from Turkish companies.

## Contribution

A novel hybrid model combining Random Forest and AdaBoost is introduced, achieving 89% accuracy in predicting going concern opinions.

## Key findings

- The hybrid model outperformed 30 traditional and other hybrid ML models in accuracy.
- The k-fold method ensured reliable results by minimizing errors from data distribution.
- The balanced dataset from Turkish companies enhanced the reliability of the research.

## Abstract

The importance of forecasting company bankruptcies makes the auditor’s reporting of the going concern opinion (GCO) a focal point for interested parties. Therefore, researchers have recently turned to predicting GCO using various machine learning (ML) methods. The aim of this research is to propose a novel hybrid model that integrates ML models to enhance the prediction accuracy of the system. We use a combination of traditional (classical) and hybrid ML approaches to identify the superior model among 30 models based on empirical data of Turkish companies listed on Borsa Istanbul (BIST) for the period 2017–2021. Given that the distribution of classes in the analysed dataset is balanced, it can be confidently stated that the research is reliable. The ML models are selected in accordance with the non-linear system since the equation system under consideration is the non-linear system. To minimise deviations and errors caused by distribution and fragmentation, the k-fold method is used to separate the training and test data sets. The experimental results show that the Random Forest based AdaBoost hybrid model outperforms traditional and other hybrid ML models in terms of accuracy by 89%.