Castroman, Nixon (2026) Laporan Kerja Praktek di [PT NTT Data Indonesia] Periode [20 Januari 2026 - 19 Juli 2026]: Refaktorisasi Pipeline Forward-Looking Probability of Default IFRS 9 dari Arsitektur Monolitik Oracle ke Arsitektur Heksagonal Berbasis Kafka. Project Report. [s.n.], [s.l.]. (Unpublished)
![[thumbnail of 5025231024-Project_Report.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5025231024-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (737kB) | Request a copy |
Abstract
Laporan kerja praktek ini membahas proses refaktorisasi pipeline Forward-Looking Probability of Default (PD) sesuai standar IFRS 9 di [PT NTT Data Indonesia]. Sistem awal (legacy) merupakan aplikasi monolitik berbasis Oracle yang sulit dipelihara dan dikembangkan. Pada kerja praktek ini, sistem tersebut direfaktorisasi menjadi arsitektur heksagonal (hexagonal architecture) berbasis pesan Apache Kafka, dengan tujuan utama menjaga paritas matematis 100% terhadap keluaran sistem legacy. Pipeline mencakup tiga algoritma perhitungan Forward-Looking PD, yaitu Normal Inversion (Binomial), Scaling, dan Merton-Vasicek. Untuk memverifikasi paritas, dibangun sebuah comparison harness yang membandingkan keluaran sistem hasil refaktorisasi terhadap sistem legacy hingga toleransi 1e-7 (tujuh angka di belakang koma). Selain itu, dilakukan optimasi pada tahap RegressionTest yang merupakan proses paling berat, melalui perhitungan closed-form, pengurangan penggunaan memori (memory streaming), serta pembatasan jumlah model keluaran agar dapat berjalan dalam batas sumber daya perangkat target (16 GB RAM, 6 CPU). Hasil akhir menunjukkan bahwa sistem hasil refaktorisasi mampu mencapai paritas matematis dengan sistem legacy sekaligus meningkatkan keterbacaan, kemudahan pemeliharaan, dan efisiensi penggunaan sumber daya.
==================================================================================================================================
This internship report discusses the refactoring of a Forward-Looking Probability of Default (PD) pipeline compliant with the IFRS 9 standard at [PT NTT Data Indonesia]. The original (legacy) system was an Oracle-based monolithic application that was difficult to maintain and extend. During this internship, the system was refactored into a hexagonal architecture driven by Apache Kafka messaging, with the primary goal of preserving 100% mathematical parity with the legacy outputs. The pipeline covers three Forward-Looking PD computation algorithms: Normal Inversion (Binomial), Scaling, and Merton-Vasicek. To verify parity, a comparison harness was built to compare the refactored outputs against the legacy system down to a tolerance of 1e-7 (seven decimal places). In addition, the RegressionTest stage—the most computationally intensive step—was optimized through closed-form computation, memory streaming to reduce RAM usage, and an output-model cap so it can run within the target machine's resource budget (16 GB RAM, 6 CPU). The final results show that the refactored system achieves mathematical parity with the legacy system while improving readability, maintainability, and resource efficiency.
| Item Type: | Monograph (Project Report) |
|---|---|
| Uncontrolled Keywords: | IFRS 9, forward-looking probability of default, refaktorisasi perangkat lunak, arsitektur heksagonal, Apache Kafka, paritas matematis, probability of default berwawasan ke depan, software refactoring, hexagonal architecture, mathematical parity, Merton-Vasicek, regresi linier |
| Subjects: | Q Science > QA Mathematics > QA76.754 Software architecture. Computer software Q Science > QA Mathematics > QA76.758 Software engineering |
| Divisions: | Faculty of Information Technology > Informatics Engineering > 55201-(S1) Undergraduate Thesis |
| Depositing User: | Nixon Castroman |
| Date Deposited: | 25 Jun 2026 07:19 |
| Last Modified: | 25 Jun 2026 07:19 |
| URI: | http://repository.its.ac.id/id/eprint/134065 |
Actions (login required)
 |
View Item |