SYSTEMATIC LITERATURE REVIEW: DATA MINING DAN ARTIFICIAL INTELLIGENCE SEBAGAI ALAT DETEKSI KECURANGAN LAPORAN KEUANGAN

SYSTEMATIC LITERATURE REVIEW: DATA MINING AND ARTIFICIAL INTELLIGENCE AS TOOLS FOR DETECTING FINANCIAL REPORTING FRAUD

Kecurangan laporan keuangan dapat merusak kepercayaan pemangku kepentingan dan mengganggu stabilitas finansial. Studi ini secara sistematis mengkaji penerapan data mining dan kecerdasan buatan (AI) dalam deteksi kecurangan laporan keuangan menggunakan pendekatan Systematic Literature Review (SLR) berdasarkan panduan (Kitchenham & Brereton, 2007). Data diperoleh dari artikel ilmiah di Scopus, ScienceDirect, dan Emerald melalui pencarian sistematis, kriteria inklusi-eksklusi, dan analisis naratif. Hasil menunjukkan bahwa Support Vector Machine (SVM) dan Logistic Regression (LR) merupakan metode yang paling banyak digunakan, sementara XGBoost dan Long Short-Term Memory (LSTM) menunjukkan performa tertinggi, masing-masing dengan akurasi sebesar 96,94% dan 94,98%. Selain itu, data keuangan seperti rasio likuiditas dan profitabilitas sering digunakan, disertai data non-keuangan seperti tata kelola perusahaan dan teks naratif dalam laporan manajemen. Temuan ini menegaskan bahwa AI dan data mining efektif meningkatkan akurasi deteksi kecurangan, serta merekomendasikan eksplorasi lebih lanjut terhadap metode unsupervised dan pemanfaatan data tidak terstruktur.

Kata kunci: kecurangan laporan keuangan, artificial intelligence, data mining, rasio keuangan, Systematic Literature Review

Financial statement fraud can undermine stakeholder trust and disrupt financial stability. This study systematically examines the application of data mining and artificial intelligence (AI) in detecting financial statement fraud using a Systematic Literature Review (SLR) approach based on the guidelines of (Kitchenham & Brereton, 2007). Data were collected from scholarly articles indexed in Scopus, ScienceDirect, and Emerald through systematic searches, inclusion–exclusion criteria, and narrative analysis. The results show that Support Vector Machine (SVM) and Logistic Regression (LR) are the most commonly used methods, while XGBoost and Long Short-Term Memory (LSTM) demonstrate the highest performance, with accuracy rates of 96.94% and 94.98%, respectively. Additionally, financial data such as liquidity and profitability ratios are frequently used, alongside non-financial data such as corporate governance indicators and narrative texts from management reports. These findings confirm that AI and data mining are effective in improving fraud detection accuracy and suggest further exploration of unsupervised methods and the utilization of unstructured data.

Keywords: financial statement fraud, artificial intelligence, data mining, machine learning, financial ratios, Systematic Literature Review.