Analisis Komparasi Klasifikasi Financial Statement Fraud pada Perusahaan Asuransi di Indonesia Menggunakan Algoritma C4.5 dan SVM dengan Pengoptimasian Bootstrap Aggregating

Adekamtari, Putri Nabila (2024) Analisis Komparasi Klasifikasi Financial Statement Fraud pada Perusahaan Asuransi di Indonesia Menggunakan Algoritma C4.5 dan SVM dengan Pengoptimasian Bootstrap Aggregating. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006201091-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (3MB) | Request a copy

Abstract

Pertumbuhan industri asuransi di Indonesia saat ini semakin baik didorong oleh faktor seperti peningkatan kesadaran masyarakat akan perlindungan finansial, pertumbuhan ekonomi yang stabil, dan regulasi yang mendukung. Namun, seiring dengan pertumbuhannya, juga muncul risiko terkait, termasuk potensi kecurangan dalam laporan keuangan. Meskipun perusahaan asuransi di Indonesia diatur oleh Otoritas Jasa Keuangan (OJK) dan harus mematuhi standar akuntansi dan pelaporan keuangan yang ketat, deteksi dini terhadap kecurangan penting dilakukan untuk menjaga integritas dan kepercayaan dalam industri ini. Pada penelitian ini dilakukan pengkajian dan membandingkan metode klasifikasi yang efektif untuk mendeteksi kecurangan (fraud) dalam laporan keuangan perusahaan asuransi di Indonesia. Dengan menggunakan penggunaan algoritma C4.5 dan Support Vector Machine (SVM) yang dioptimalkan dengan metode Bootstrap Aggregating (Bagging) berdasarkan nilai dari Beneish M-score Model serta faktor penyebab fraud pada laporan keuangan yang dianalisis dengan pendekatan fraud hexagon theory. Berdasarkan hasil klasifikasi, Model Algoritma C4.5 berhasil mengklasifikasikan financial statement fraud dengan akurasi 62,42%, dengan nilai precision 70,26%, dan recall 60,89%. Sementara itu, SVM polynomial menunjukkan performa unggul dengan nilai akurasi 71,37%, precision 77,74%, dan recall 71,61%, menunjukkan efektivitasnya sebagai klasifikasi tunggal. Pengoptimalan bagging pada Algoritma C4.5 meningkatkan akurasi menjadi 67,88%, precision menjadi 73,92%, dan recall menjadi 71,43%, menunjukkan peningkatan kinerja dalam mendeteksi fraud. Namun, optimasi bagging pada SVM menurunkan akurasi menjadi 68,35%, meskipun recall meningkat menjadi 72,86%, menunjukkan peningkatan sensitivitas terhadap fraud tetapi dengan lebih banyak false positives. Secara keseluruhan, SVM polynomial tetap menunjukkan performa terbaik dalam mengidentifikasi financial statement fraud pada perusahaan asuransi di Indonesia, meskipun hanya sebagai klasifikasi tunggal.

============================================================

The growth of the insurance industry in Indonesia is currently improving, driven by factors such as increasing public awareness of financial protection, stable economic growth, and supportive regulations. However, along with its growth, there are also emerging related risks, including the potential for fraud in financial reporting. Although insurance companies in Indonesia are regulated by the Otoritas Jasa Keuangan (OJK) and must comply with strict accounting and financial reporting standards, early detection of fraud is important to maintain integrity and trust in this industry. This research assesses and compares effective classification methods for detecting fraud in the financial statements of insurance companies in Indonesia. By using the C4.5 algorithm and Support Vector Machine (SVM) optimized with the Bootstrap Aggregating (Bagging) method based on the values of the Beneish M-score Model, as well as the factors causing fraud in financial statements analyzed with the fraud hexagon theory approach. Based on the classification results, the C4.5 algorithm model successfully classified financial statement fraud with an accuracy of 62.42%, a precision of 70.26%, and a recall of 60.89%. Meanwhile, the polynomial SVM model showed superior performance with an accuracy of 71.37%, a precision of 77.74%, and a recall of 71.61%, demonstrating its effectiveness as a single classifier. Bagging optimization on the C4.5 algorithm increased the accuracy to 67.88%, precision to 73.92%, and recall to 71.43%, showing improved performance in detecting fraud. However, bagging optimization on the SVM decreased accuracy to 68.35%, although recall increased to 72.86%, indicating increased sensitivity to fraud but with more false positives. Overall, the polynomial SVM model still showed the best performance in identifying financial statement fraud in insurance companies in Indonesia, even as a single classifier.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Algoritma C4.5, Beneish M-Score, Bootstrap Aggregating, C4.5 Algorithm, Fraud Hexagon Theory, Support Vector Machine
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Putri Nabila Adekamtari
Date Deposited:	01 Aug 2024 02:02
Last Modified:	01 Aug 2024 02:02
URI:	http://repository.its.ac.id/id/eprint/110002

Actions (login required)

View Item