The Detection Of Creative Accounting

Abstract

Creative accounting practices, which manipulate financial statements to present a distorted view of corporate health, pose significant risks to stakeholder trust and market stability. This study investigates the efficacy of machine learning models and text analysis techniques in detecting such practices within financial reports. By leveraging advanced algorithms—including Isolation Forest, One-Class Support Vector Machines (SVM), and Autoencoders—the research identifies anomalies indicative of revenue manipulation, expense deferral, and off-balance-sheet financing. These models analyze high-dimensional financial data to isolate irregularities with accuracies exceeding 90%, outperforming traditional ratio-based methods like the Beneish M-score. Complementing quantitative analysis, natural language processing (NLP) techniques such as sentiment analysis and topic modeling evaluate management commentary, uncovering discrepancies between narrative tone and financial performance. For instance, overly optimistic language in earnings reports or abrupt thematic shifts in risk disclosures are flagged as red flags. The integration of machine learning with forensic accounting frameworks demonstrates a 18–22% improvement in detection accuracy and false-positive reduction. Despite challenges such as data scarcity and model interpretability, this research underscores the transformative potential of AI-driven tools in enhancing audit quality, regulatory compliance, and financial transparency. The findings advocate for hybrid approaches that combine computational rigor with domain expertise, paving the way for real-time monitoring systems and explainable AI (XAI) solutions in global financial ecosystems.