Automated Auditing Based on Machine Learning: Model Construction and Empirical Analysis

Abstract

This study explores how machine learning (ML) brings automation to auditing as a way to boost accuracy as well as efficiency and detect fraud. An evaluation of financial anomaly detection was conducted using decision trees, random forests, XGBoost, and neural networks as the tested machine learning models. The analytical models evaluated Neural Networks for the most accurate predictions, although XGBoost and Random Forest offered optimal accuracy versus computational performance. The selected financial indicators consisted of Total Revenue and Net Profit Margin, which proved instrumental for assessment purposes. The implementation of these benefits requires solving issues regarding interpretability along with bias concerns and compliance regulations. The research exhibits how machine learning technologies can transform auditing by demonstrating the necessity to advance these technological systems for better audit transparency and reliability in financial inspections.