Pengembangan Metode Proses Discovery Berbasis Graf Dengan Memperhatikan Waktu Aktivitas Pada Proses Mengandung Relasi Paralel

Pujangga, Dimas Aria (2025) Pengembangan Metode Proses Discovery Berbasis Graf Dengan Memperhatikan Waktu Aktivitas Pada Proses Mengandung Relasi Paralel. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025211212-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Model proses berperan penting bagi analisis untuk memahami dan menganalisis alur kerja bisnis. Disebutkan bahwa dari berbagai metode process discovery yang ada, metode time-based heuristic miner menerapkan ambang batas relasi yang ditampilkan supaya model proses yang sulit dianalisa karena variasi trace yang banyak (model proses spaghetti) menjadi model proses yang lebih mudah dianalisa (model proses lasagna). Selain itu, metode time-based heuristic miner memperhatikan waktu awal dan waktu akhir aktivitas supaya dapat menggambarkan relasi paralel, yaitu OR dan AND, dari event log yang tidak lengkap kombinasi trace-nya. Namun, metode time-based heuristic miner memiliki beberapa kekurangan yaitu tidak ada aplikasi tunggal yang mendukung penyimpanan data dan visualisasi model proses, dan belum dapat menangani relasi tingkat lanjut seperti invisible tasks. Penelitian ini mengusulkan metode process discovery berbasis graf dengan memperhatikan waktu aktivitas untuk membangun model proses berbasis graf berdasarkan event log yang tidak lengkap kombinasi trace-nya, menggunakan graph database untuk menyimpan dan menampilkan event log maupun model proses dalam satu aplikasi, serta membentuk rule berbasis graf untuk menggambarkan invisible task pada model proses. Evaluasi penelitian ini menggunakan tiga dataset, yaitu dataset pertama yang memiliki 14 traces, dataset kedua yang memiliki 11 traces, dan dataset ketiga yang memiliki 1290 traces. Pada dataset pertama dan kedua, hasil evaluasi menunjukkan metode process discovery penelitian ini, yaitu metode time-based graph process discovery menghasilkan rata-rata fitness 0,879, precision 0,52, generalization 0,865, dan simplicity 1, sedangkan metode heuristic miner menghasilkan hasil yang sama yaitu fitness 0,879, precision 0,52, generalization 0,865, dan simplicity 1. Pada dataset ketiga, hasil evaluasi menunjukkan metode time-based graph process discovery menghasilkan simplicity 1, sedangkan metode heuristic miner juga menghasilkan simplicity 1.
======================================================================================================================================
Process models play a crucial role in enabling analysis to understand and evaluate business workflows. By leveraging data from event logs, process discovery constructs relationships between activities to generate a process model that clearly illustrates the workflow. Among the various process discovery methods available, the time-based heuristic miner applies a threshold on the relationships it visualizes in order to simplify complex process models caused by high trace variability (commonly referred to as spaghetti models) into more analyzable models (lasagna models). In addition, the time-based heuristic miner takes into account the start and end times of activities to identify parallel relations—namely OR and AND—even when the event log lacks complete trace combinations. However, the time-based heuristic miner has several limitations. It lacks a unified application for both data storage and process model visualization, and it is not capable of handling advanced relationships such as invisible tasks. This study proposes a graph-based process discovery method that considers activity timing to construct graph-based process models using event logs with incomplete trace combinations. It also utilizes a graph database to store and visualize both log data and process models within a single application and defines graph-based rules to represent invisible tasks in the process model. The evaluation in this study uses three datasets: the first dataset containing 14 traces, the second dataset with 11 traces, and the third with 1,290 traces. For the first and second datasets, the evaluation results show that the proposed time-based graph process discovery method achieved average scores of fitness 0.879, precision 0.52, generalization 0.865, and simplicity 1. The heuristic miner method produced identical results. For the third dataset, the evaluation revealed that both the proposed method and the heuristic miner yielded simplicity of 1.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Model Proses Bisnis Berbasis Graf, Invisible Tasks, Process Discovery, Time-Based Heuristic Miner, Graph-Based Modeling, Invisible Tasks, Process Discovery, Time-Based Heuristic Miner.
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Dimas Aria Pujangga
Date Deposited:	29 Jul 2025 05:36
Last Modified:	29 Jul 2025 05:36
URI:	http://repository.its.ac.id/id/eprint/122589

Actions (login required)

View Item