Pengembangan metode deteksi anomali menggunakan modifikasi dan metode ensemble: penerapan pada laporan keuangan Indonesia = Improvement of anomaly detection methods using modification and ensemble method: application in Indonesian financial statement

Putri, Vanessa Laurencia Hartoyo (2023) Pengembangan metode deteksi anomali menggunakan modifikasi dan metode ensemble: penerapan pada laporan keuangan Indonesia = Improvement of anomaly detection methods using modification and ensemble method: application in Indonesian financial statement. Bachelor thesis, Universitas Pelita Harapan.

	Text (Title) Title.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (254kB)
	Text (Abstract) Abstract.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (1MB)
	Text (ToC) ToC.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (622kB)
	Text (Chapter1) Chapter1.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (583kB)
	Text (Chapter2) Chapter2.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (864kB)
	Text (Chapter3) Chapter3.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (821kB)
	Text (Chapter4) Chapter4.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (1MB)
	Text (Chapter5) Chapter5.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (574kB)
	Text (Bibliography) Bibliography.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (572kB)
	Text (Appendices) Appendices.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (3MB)

Abstract

Dalam menunjukkan citra perusahaan yang baik terhadap publik, perusahaan dapat melakukan manipulasi pada laporan keuangan. Penelitian ini bertujuan untuk menemukan metode machine learning yang dapat mendeteksi manipulasi melalui anomali yang ada pada laporan keuangan. Deteksi anomali akan dilakukan dengan perancangan model dengan menggunakan jarak Mahalanobis, Local Outlier Factor, Isolation Forest, dan pengembangan model akan dilakukan dengan menggunakan jarak Mahalanobis - Minimum Covariance Determinant, Isolation Forest - Local Outlier Factor, dan metode ensemble dengan menggunakan majority voting. Hasil dari deteksi anomali akan dibandingkan dengan Beneish M-score untuk melihat hubungan dari kedua model tersebut. Nilai yang didapatkan akan dibandingkan untuk menemukan metode deteksi anomali terbaik. Hasil penelitian menunjukkan bahwa model deteksi anomali masih memiliki hubungan yang lemah dengan model Beneish M-score. Perbandingan antar model juga menunjukkan bahwa hampir tidak ada perbedaan yang signifikan antara model dengan pengembangannya. / In order to show a good image to the public, a company could manipulate its financial statement. This research aims to find a machine learning method that could detect manipulations through anomalies inside the financial statement. Anomaly detection will be done by designing models using Mahalanobis distance, Local Outlier Factor, Isolation Forest, and the development of the models will be done by using Mahalanobis distance - Minimum Covariance Determinant, Isolation Forest - Local Outlier Factor, and majority voting. Results from the anomaly detection will be compared with Beneish M-score to see the relationship between the models. The outcome will be used to find the best anomaly detection method. The result shows that the relationship between the anomaly detection models and Beneish M-score is not strong. The comparison between the models shows that there is almost no significant difference between the models and its development.

Actions (login required)

View Item