Abstract: Financial statement misstatement remains a persistent audit challenge because misstatement risk is rarely distributed uniformly across account balances, journal entries, or transaction populations. Conventional audit sampling provides a disciplined basis for obtaining audit evidence, but it may under-detect misstatements when fraud is strategically concealed, when error clusters in high-complexity accounts, or when nonlinear interactions exist among accruals, revenue growth, control weaknesses, manual entries, and materiality exposure. This paper develops an adaptive audit risk scoring algorithm for detecting financial statement misstatements using machine learning and compares the method with traditional sampling techniques under equivalent audit effort. The proposed Adaptive Audit Risk Score (ARS) integrates calibrated misstatement probability, anomaly intensity, materiality exposure, control-risk score, and model uncertainty. A simulation-based audit population is used to evaluate logistic regression, random forest, gradient boosting, support vector machine, and an adaptive hybrid ARS against simple random sampling, systematic sampling, stratified sampling, and monetary-unit sampling. The methodology is rooted in the audit risk model, probabilistic classification, anomaly detection, class-weighted loss functions, evidence updating, and cost-aware threshold optimization. Results from the simulated benchmark show that adaptive scoring concentrates audit effort on higher-risk items and detects substantially more misstated observations than conventional sampling at the same top 10% testing budget. The study contributes to audit theory by translating the classical audit risk model into a dynamic probability-updating framework, and it contributes to practice by offering a defensible, explainable, and materiality-aware procedure for risk-based audit selection. The paper concludes that machine learning should not replace professional judgement, but it can materially improve sampling prioritization, substantive testing design, fraud-risk response, and audit-cost efficiency when embedded in an appropriate governance and validation structure.
Keywords: adaptive audit risk, financial statement misstatement, machine learning, audit sampling, anomaly detection, materiality, probability calibration, audit analytics.
Title: Adaptive Audit Risk Scoring Algorithm for Detecting Financial Statement Misstatements Using Machine Learning and Comparative Analysis with Traditional Sampling Techniques
Author: Sallah Joseph Kadir, Rukayat Akingbade, Otugene Victor Bamigwojo, Joy Onma Enyejo
International Journal of Recent Research in Commerce Economics and Management (IJRRCEM)
ISSN 2349-7807
Vol. 13, Issue 3, July 2026 - September 2026
Page No: 1-21
Paper Publications
Website: www.paperpublications.org
Published Date: 01-July-2026
