Wahyuni, Cahyaningtyas Sekar (2020) Conditional Causal Footprint Matching Untuk Deteksi Anomali Illegal Pattern Dan Wrong Decision Pada Transaksi Enterprise Resource Planning (ERP). Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
05111850010005-Master_Thesis.pdf - Accepted Version Download (4MB) | Preview |
Abstract
Enterprise Resource Planning (ERP) adalah aplikasi yang digunakan untuk mendukung proses bisnis. Namun, tidak semua proses sepenuhnya otomatis. Dengan demikian, kondisi ini menyebabkan munculnya proses anomali. Beberapa metode anomali, yaitu Process-based Fraud dan Anomaly Pattern Rule, masih memiliki kelemahan dalam mendeteksi anomali. Metode tersebut gagal dalam menangkap illegal pattern karena metode tersebut hanya mempertimbangkan urutan kegiatan tanpa memperhatikan konten dan informasi tambahan yang terkait dengan kegiatan. Penelitian ini mengusulkan metode yang mengombinasikan conditional causal footprint, metode similaritas baru, dan algoritma pattern matching untuk menangkap anomali. Conditional causal footprint memperluas causal footprint dengan menambahkan nilai kondisional dari suatu aktivitas. Selanjutnya, conditional causal footprint diukur dengan metode similaritas baru untuk mendapatkan hasil similaritas tertinggi dari setiap trace. Penelitian ini mengusulkan metode similaritas untuk memberikan hasil yang lebih tepat dengan membedakan operator hubungan paralel, yaitu AND, OR, dan XOR. Terakhir, hasil tertinggi dari metode similaritas dibandingkan melalui pattern matching untuk menangkap anomali. Hasil eksperimen menunjukkan bahwa metode usulan mampu meningkatkan sensitivitas, spesifisitas, dan akurasi hingga 100%. Hal ini menunjukkan bahwa metode usulan lebih baik dalam mendeteksi anomali dibandingkan metode terdahulu.
=================================================================================================================================
Enterprise Resource Planning (ERP) is an application used to support business processes of a company. Nevertheless, in practical not all business processes are fully-automated. Thus, this condition led to the emergence of anomalous processes. Several existing anomaly methods, i.e. Process-based Fraud and Anomaly Pattern Rules, still have drawbacks in detecting anomaly. Firstly, the thirteen types of anomaly (or can be called anomaly attributes) failed to capture
illegal patterns. Secondly, they only considered the activities order without paying attention to the content and additional information related to the activities. This research proposes a method that combined conditional causal footprint, novel similarity method, anSd pattern matching algorithm to capture the anomaly. Conditional causal footprint extends causal footprint by adding conditional value of the activities. Next, the conditional causal footprints are measured by the novel similarity method to obtain the highest similarity result of each trace. This research proposes a similarity method to give a more precise result by differentiating the operator of parallel relationships, i.e. AND, OR, and XOR. Lastly, the highest result of similarity method is compared through pattern matching to capture the anomalies. The experiment results showed that the proposed method was able to increase the sensitivity, specificity, and accuracy to 100% each. It can be concluded that the proposed method was able to detect anomaly process compared to other previous methods.
| Item Type: | Thesis (Masters) |
|---|---|
| Additional Information: | RTIf 658.401 2 Wah c-1 |
| Uncontrolled Keywords: | anomaly detection, behavioral similarity, conditional causal footprint, illegal pattern, pattern matching |
| Subjects: | T Technology > T Technology (General) > T58.6 Management information systems |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis |
| Depositing User: | Cahyaningtyas Sekar Wahyuni |
| Date Deposited: | 11 Mar 2025 03:32 |
| Last Modified: | 11 Mar 2025 03:33 |
| URI: | http://repository.its.ac.id/id/eprint/73224 |
Actions (login required)
 |
View Item |