INTEGRATED ACCOUNTING AND FORENSIC AUDITING FOR COMPLEX ENGINEERING PROJECTS: A DATA-DRIVEN FRAMEWORK FOR DETECTING COST MANIPULATION, PROCUREMENT FRAUD, AND CONTRACTUAL IRREGULARITIES

Abstract

Complex engineering projects generate large volumes of interconnected financial, procurement, technical, and contractual records, making conventional periodic audits insufficient for identifying sophisticated irregularities. This study develops an integrated accounting and forensic auditing framework for detecting cost manipulation, procurement fraud, and contractual irregularities throughout the engineering project lifecycle. The investigation employed a three-year anonymized dataset containing 285,000 financial and operational records from 24 complex projects across the construction, energy, transportation, manufacturing, and public-infrastructure sectors. The records covered invoices, purchase orders, supplier quotations, tender evaluations, payment certificates, variation orders, payroll entries, equipment logs, progress reports, contract amendments, and internal-audit observations. After data cleaning, reconciliation, normalization, categorical encoding, and duplicate removal, 46 financial, procurement, contractual, temporal, and organizational risk indicators were constructed. Independent forensic verification identified 13,965 irregular records, representing 4.9% of the final dataset. The proposed framework combined forensic ratio analysis, Benford’s Law testing, process mining, supplier-network analysis, document matching, and machine-learning classification. Its performance was compared with logistic regression, support vector machine, random forest, XGBoost, LightGBM, and artificial neural network models using a leakage-free temporal data split. The integrated framework achieved 95.6% accuracy, 93.8% precision, 92.7% recall, a 93.2% F1-score, and an area under the receiver operating characteristic curve of 0.974. It reduced false-positive alerts by 31.4% and detected suspicious activities an average of 18 days earlier than conventional manual auditing. Major cost-manipulation patterns included inflated quantities, duplicate billing, concealed variation orders, and premature cost recognition. Procurement fraud was associated with bid rotation, supplier collusion, split purchasing, abnormal price deviations, and undisclosed conflicts of interest, while contractual irregularities involved unauthorized amendments, payment–progress mismatches, repeated threshold exemptions, and unsupported claims. It further connects flagged transactions to responsible personnel, suppliers, approval paths, and project milestones, thereby supporting targeted investigations and defensible audit conclusions in practice. The findings demonstrate that integrating accounting controls with forensic analytics strengthens audit coverage, evidential traceability, and detection timeliness. The framework offers project owners, auditors, regulators, and engineering managers a practical risk-based mechanism for prioritizing investigations, limiting financial leakage, improving procurement transparency, and reinforcing accountability in complex engineering projects.