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

Area of Science:

• Accounting
• Computer Science
• Financial Risk Management

Background:

• Forecasting company bankruptcies is crucial for stakeholders.
• Auditor's going concern opinion (GCO) is a key indicator.
• Machine learning (ML) methods are increasingly used for GCO prediction.

Purpose of the Study:

• To propose a novel hybrid ML model for enhanced GCO prediction accuracy.
• To compare 30 traditional and hybrid ML models using empirical data.
• To identify the superior ML model for predicting GCO.

Main Methods:

• Utilized empirical data from Turkish companies listed on Borsa Istanbul (2017-2021).
• Employed a combination of traditional and hybrid ML approaches.
• Used k-fold cross-validation for robust model evaluation on a balanced dataset.

Main Results:

• The Random Forest based AdaBoost hybrid model demonstrated superior performance.
• This hybrid model achieved an accuracy of 89%, outperforming other models.
• The study confirmed the reliability of results due to a balanced dataset.

Conclusions:

• Hybrid ML models, particularly Random Forest based AdaBoost, offer significant improvements in GCO prediction.
• The findings provide valuable insights for auditors and investors in assessing financial risk.
• This research contributes to the advancement of ML applications in financial forecasting.