Metode Pencarian Kembali Proses Hilang Dalam Model Proses Bisnis Menggunakan Repositori

Kristina, Kristina (2026) Metode Pencarian Kembali Proses Hilang Dalam Model Proses Bisnis Menggunakan Repositori. Doctoral thesis, INSTITUT TEKNOLOGI SEPULUH NOVEMBER.

Text 7025221016-Doctoral.pdf Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Menghasilkan model proses bisnis (Business Process Model, BPM) dari dokumen informal memiliki risiko, seperti hilangnya proses dan/atau aliran dalam model tersebut. Hal ini disebabkan oleh narasumber yang diwawancarai hanya mengandalkan ingatan dan kebiasaan sehari-hari dalam mengelola operasional organisasi, sehingga beberapa proses mungkin terlewat. Hasil wawancara kemudian diubah menjadi diagram BPMN menggunakan notasi Business Process Model and Notation (BPMN). Diagram BPMN yang dihasilkan dari dokumen informal tersebut diubah menjadi graf interstruktural dan intrastruktural. Graf hasil translasi dokumen informal memiliki risiko kehilangan simpul proses dan/atau aliran.
Penelitian ini telah membangun metode pencarian proses dan\ atau aliran yang hilang (Repolink) secara otomatis di BPM dengan mengkombinasikan pengukuran keserupaan struktural dan semantik berbasis repositori secara efisien. Repolink menggunakan lima algoritma seperti Graph Edit Distance (GED), Greedy, Cosine Similarity, Algoritma Kahn, dan Breadth First Search (BFS) untuk mengukur kesamaan struktural dan semantik graf. Repolink menekankan efisiensi komputasi dengan pengadaan mekanisme penghentian dini, yang memungkinkan identifikasi simpul atau tepi yang hilang secara cepat. Jika ada simpul yang nilai keserupaan struktural dan semantik tidak terpenuhi, maka graf tersebut akan dieliminasi dari pencarian. Pencarian akan dilanjutkan pada graf yang memenuhi nilai keserupaan minimal. Metode ini efisien karena menghemat waktu pencarian. Pencarian tidak perlu menghitung nilai keserupaan satu graf BPM dengan semua graf BPM pada repositori secara menyeluruh.
Repolink berhasil mengidentifikasi dan mengkontruksi proses dan/atau aliran yang hilang dari BPM dengan nilai kesepakatan pakar 0,93 yang berarti metode dapat diandalkan. Repolink menunjukkan performa yang efisien terhadap peningkatan ukuran dataset di repositori setelah diukur dengan Algoritma Big O yaitu O (B * A). Efisiensi Repolink diperoleh dengan menerapkan mekanisme early stopping pada simpul yang tidak memenuhi nilai keserupaan minimal yang telah ditentukan. Hal ini terjadi karena Repolink bekerja berdasarkan hubungan langsung antar simpul dari graf.
==================================================================================================================================
Generating a Business Process Model (BPM) from informal documents carries risks, such as omitting processes or flows. This occurs because interviewees rely solely on memory and daily routines for managing organizational operations, which can lead to overlooked processes. Interview findings are then converted into BPMN diagrams using Business Process Model and Notation (BPMN) standards. These diagrams are subsequently transformed into inter-structural and intrastructural graphs. Graphs resulting from this translation risk losing process nodes and/or flows. This research developed Repolink, a method that first identifies missing processes and/or flows in BPMs by combining repository-based structural and semantic similarity measurements. Repolink then applies five algorithms, such as Graph Edit Distance (GED), Greedy, Cosine Similarity, Kahn's Algorithm, and Breadth-First Search (BFS), to assess graph structural and semantic similarity. The method uses an early-stopping mechanism to identify missing nodes or edges rapidly. Graphs that do not meet the required structural and semantic similarity thresholds are eliminated from the search, while the search continues with graphs that satisfy the minimum similarity criteria. This method is efficient, as it reduces search time and eliminates the need for an exhaustive comparison of a single BPM graph against every other BPM graph in the repository. Repolink successfully identified and reconstructed missing processes and/or flows in BPMs, achieving an expert agreement score of 0.93, which indicates the method's reliability. Performance analysis using Big O notation, specifically O(B*A), demonstrates that Repolink remains efficient as the repository dataset grows. This efficiency is achieved by applying an early-stopping mechanism to nodes that fail to meet the predetermined minimum similarity threshold, before continuing with nodes supported by Repolink's reliance on direct relationships between graph nodes.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	pencarian informasi, keserupaan graf, struktural semantik, graph edit distance (GED), algoritma greedy.information retrieval, graph similarity, structural semantics, graph edit distance (GED), greedy
Subjects:	T Technology > T Technology (General) > T58.6 Management information systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55001-(S3) PhD Thesis (Comp Science)
Depositing User:	Kristina Kristina
Date Deposited:	04 Aug 2026 07:16
Last Modified:	04 Aug 2026 07:16
URI:	http://repository.its.ac.id/id/eprint/138614

Actions (login required)

View Item