The research matters as Indonesia’s financial sector increasingly relies on digital platforms. Cooperatives, fintech services, and small-to-medium enterprises now process thousands of transactions daily through automated systems. While this digital shift improves efficiency, it also increases exposure to fraud that can slip through traditional, manual auditing processes. The study demonstrates that artificial intelligence can play a critical role in closing this gap by detecting behavioral patterns linked to fraud early and consistently.
Why Digital Fraud Detection Needs a New Approach
Across the global digital economy, financial transactions are growing in volume, speed, and complexity. In Indonesia, digital cooperatives have become essential financial service providers, especially in local communities. Many of these institutions manage large numbers of small, repetitive transactions—conditions that make manual monitoring difficult and time-consuming.
Traditional audits often rely on sampling and periodic checks. These methods can miss small irregularities that accumulate into major losses. The authors argue that this challenge is especially acute in digital cooperatives, where transactions occur around the clock and internal control systems are often limited.
Artificial intelligence offers a practical response. By processing large datasets in real time and identifying abnormal transaction patterns, AI can support forensic accountants and auditors in spotting risks that humans might overlook.
How the Researchers Studied AI in Forensic Accounting
The research analyzed 2,314 anonymized digital transaction records from a technology-based savings and loan cooperative in Central Java over one year. All data were extracted directly from the cooperative’s accounting information system and anonymized to protect member privacy.
Instead of relying on a single analytical tool, the researchers tested three widely used machine learning models:
· Random Forest
· Logistic Regression
· Support Vector Machine (SVM)
The models were trained on 80 percent of the transaction data and tested on the remaining 20 percent. To ensure fair results, the team balanced the dataset so that rare fraud cases were adequately represented. Model performance was evaluated using accuracy, precision, recall, and related metrics commonly used in fraud analytics.
Key Findings at a Glance
The results show that artificial intelligence significantly improves fraud detection compared with rule-based or manual systems.
Main findings include:
· Random Forest achieved the highest accuracy at 91.8 percent, outperforming logistic regression and SVM.
· Around 7.4 percent of transactions showed characteristics associated with potential fraud.
· Fraud detection was driven more by behavioral patterns than by transaction size alone.
The analysis identified three variables that consistently signaled higher fraud risk:
· High transaction frequency
· Repeated late payments
· Frequent loan applications
“These indicators reveal behavioral anomalies rather than isolated financial values,” the authors explain, emphasizing that fraud often emerges from patterns that deviate from normal transaction behavior over time.
What This Means for Financial Institutions and Policymakers
The findings have clear implications for financial governance in the digital economy. For cooperatives and other non-bank institutions, AI-based forensic accounting systems can:
· Strengthen internal controls through continuous monitoring
· Reduce reliance on manual audits that are prone to human error
· Identify suspicious behavior earlier, limiting potential financial losses
For policymakers and regulators, the study supports the case for encouraging AI adoption in financial supervision frameworks, particularly for smaller institutions that lack extensive auditing resources.
Ethically, the authors stress that AI should support—not replace—professional judgment. “AI-based forensic systems need to operate within strong governance and ethical frameworks,” Ramandei and colleagues note, pointing out that human oversight remains essential in interpreting and acting on flagged transactions.
Academic Insight from the Authors
Reflecting on the broader significance of the findings, the research team highlights the role of behavioral data in modern fraud detection. An ethical paraphrase of their conclusion states that machine learning allows forensic accounting to move beyond static rules and focus on dynamic transaction behavior, which is more reflective of real-world fraud risks, particularly in digital financial environments.
This perspective reinforces the idea that AI-driven models are not just technical tools, but strategic assets for improving transparency and accountability.
Author Profile
Pilipus Ramandei, PhD Lecturer and researcher in accounting and forensic auditing at Universitas Ottow Geissler Papua, Indonesia. His expertise includes forensic accounting, financial ethics, and the application of artificial intelligence in auditing.
Kristanti Rahman, MAcc Accounting scholar at STIE Muhammadiyah Cilacap, Indonesia, specializing in financial analysis, digital accounting systems, and fraud detection.
Dharma Widada, PhD Senior academic at Universitas Mulawarman, Indonesia, with expertise in accounting information systems, risk management, and financial governance.
Source
Article Title: Artificial Intelligence Integration in Forensic Accounting for Detecting Financial Fraud in the Digital Economy
Journal: Formosa Journal of Science and Technology
Year: 2026
0 Komentar