Prediksi Klasifikasi Financial Statement Fraud Berbasis Beneish M-Score Perusahaan Industri Keuangan di Indonesia dengan Fuzzy Regresi Logistik

Eriyanti, Marshanda Geometrika (2026) Prediksi Klasifikasi Financial Statement Fraud Berbasis Beneish M-Score Perusahaan Industri Keuangan di Indonesia dengan Fuzzy Regresi Logistik. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221036-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Financial statement fraud merupakan salah satu bentuk kecurangan yang dapat menimbulkan kerugian besar serta mengganggu stabilitas perusahaan dan sistem keuangan, khususnya pada industri keuangan. Deteksi financial statement fraud tidak selalu bersifat tegas karena terdapat kondisi yang mengandung ketidakpastian sehingga sulit diklasifikasikan secara pasti sebagai fraud atau non-fraud. Penelitian ini bertujuan membangun model klasifikasi financial statement fraud menggunakan Regresi Logistik Biner dan Fuzzy Logistic Regression serta membandingkan kedua metode dengan dan tanpa penerapan Synthetic Minority Oversampling Technique (SMOTE) pada perusahaan industri keuangan yang terdaftar pada papan pencatatan utama Bursa Efek Indonesia periode 2020-2024. Variabel independen yang digunakan meliputi Financial Target (ROA), Financial Stability (ACHANGE), Director Change, Marginal Cost, Quality of External Auditor, Nature of Industry, Auditor Change, dan CEO Duality, sedangkan variabel dependen ditentukan menggunakan Beneish M-Score. Evaluasi model dilakukan menggunakan Stratified K-Fold Cross Validation dengan K=5 dan K=10 berdasarkan metrik Accuracy, Precision, Recall, Specificity, Area Under Curve (AUC), dan Mean Degree of Membership (MDM). Hasil penelitian menunjukkan bahwa model Regresi Logistik Biner terbaik diperoleh pada full model dengan Stratified 10-Fold Cross Validation yang menghasilkan Accuracy sebesar 0,761 dan AUC sebesar 0,846. Model Fuzzy Logistic Regression dengan SMOTE menggunakan skenario fuzzifikasi fraud (0,70; 0,95) dan non-fraud (0,05; 0,30) dipilih sebagai model akhir karena memberikan keseimbangan performa klasifikasi yang lebih baik berdasarkan Accuracy sebesar 0,750, Precision sebesar 0,727, Recall sebesar 0,800, Specificity sebesar 0,700, AUC sebesar 0,800, dan MDM sebesar 0,760. Interpretasi Odds Ratio menunjukkan bahwa Financial Stability merupakan satu-satunya variabel yang berpengaruh signifikan terhadap financial statement fraud, sedangkan Financial Target memiliki nilai Odds Ratio terbesar sehingga menunjukkan peningkatan risiko paling tinggi dibandingkan variabel prediktor lainnya. Hasil penelitian menunjukkan bahwa kombinasi Fuzzy Logistic Regression dan SMOTE dapat menjadi alternatif metode klasifikasi yang efektif untuk mendeteksi financial statement fraud sekaligus merepresentasikan ketidakpastian prediksi.
=====================================================================================================================================
Financial statement fraud is a form of corporate fraud that can cause substantial losses and threaten the stability of companies and the financial system, particularly in the financial industry. Detecting financial statement fraud is inherently uncertain because companies cannot always be clearly classified as fraudulent or non-fraudulent. This study aims to develop financial statement fraud classification models using Binary Logistic Regression and Fuzzy Logistic Regression and to compare the performance of both methods with and without the application of the Synthetic Minority Oversampling Technique (SMOTE) on financial sector companies listed on the Main Board of the Indonesia Stock Exchange during the 2020–2024 period. The independent variables consist of Financial Target (ROA), Financial Stability (ACHANGE), Director Change, Marginal Cost, Quality of External Auditor, Nature of Industry, Auditor Change, and CEO Duality, while the dependent variable is determined using the Beneish M-Score as a proxy for financial statement fraud. Model performance was evaluated using Stratified K-Fold Cross Validation with K=5 and K=10 based on Accuracy, Precision, Recall, Specificity, Area Under the Curve (AUC), and Mean Degree of Membership (MDM). The best Binary Logistic Regression model was obtained from the full model with Stratified 10-Fold Cross Validation, achieving an Accuracy of 0.761 and an AUC of 0.846. Meanwhile, the Fuzzy Logistic Regression model combined with SMOTE using the fraud fuzzification scenario (0.70, 0.95) and non-fraud scenario (0.05, 0.30) was selected as the final model, achieving an Accuracy of 0.750, Precision of 0.727, Recall of 0.800, Specificity of 0.700, AUC of 0.800, and MDM of 0.760. The Odds Ratio analysis indicates that Financial Stability is the only variable with a statistically significant effect on financial statement fraud, whereas Financial Target has the largest Odds Ratio, indicating the highest relative increase in fraud risk among the predictor variables. These findings suggest that the combination of Fuzzy Logistic Regression and SMOTE provides an effective alternative for detecting financial statement fraud while simultaneously representing prediction uncertainty through fuzzy membership functions.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Beneish M-Score, Financial Statement Fraud, Fuzzy Regresi Logistik, Fraud Hexagon Theory, Regresi Logistik Biner, Binary Logistic Regression, Beneish M-Score, Financial Statement Fraud Detection, Fuzzy Logistic Regression, Fraud Hexagon Theory.
Subjects:	H Social Sciences > HA Statistics > HA31.3 Regression. Correlation. Logistic regression analysis. H Social Sciences > HA Statistics > HA31.7 Estimation H Social Sciences > HC Economic History and Conditions > HC441 Macroeconomics.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Marshanda Geometrika Eriyanti
Date Deposited:	01 Aug 2026 01:31
Last Modified:	01 Aug 2026 01:31
URI:	http://repository.its.ac.id/id/eprint/139729

Actions (login required)

View Item