Process Discovery Untuk Streaming Event Log Menggunakan Model Markov Tersembunyi

Sungkono, Kelly Rossa (2016) Process Discovery Untuk Streaming Event Log Menggunakan Model Markov Tersembunyi. Undergraduate thesis, Institut Teknology Sepuluh Nopember.

[img] Text
5112100199-undergraduate theses.pdf - Accepted Version

Download (2MB)
[img] Text
5112100199-paperpdf.pdf - Accepted Version

Download (662kB)
[img] Text
5112100199-presentationpdf.pdf - Presentation

Download (1MB)

Abstract

Process discovery adalah teknik penggalian model proses dari rangkaian aktivitas yang tercatat dalam event log. Saat ini, sistem informasi menghasilkan streaming event log dimana event log dicatat sesuai waktu proses yang terjadi. Online Heuristic Miner adalah algoritma process discovery yang mampu menghasilkan model proses dari streaming event log. Kelemahan dari algoritma Online Heuristic Miner adalah ketidakmampuan mengatasi incomplete trace pada streaming event log. Incomplete trace adalah rangkaian aktivitas pada event log yang terpotong di bagian awal ataupun di bagian akhir. Incomplete trace mengakibatkan proses secara utuh tidak dapat ditampilkan dalam model proses sehingga incomplete trace disebut sebagai noise. Algoritma yang memanfaatkan Model Markov Tersembunyi digunakan untuk membentuk model proses yang dapat menangani incomplete trace pada streaming event log. Model Markov Tersembunyi dipilih karena model ini banyak digunakan untuk meramalkan data sehingga berguna dalam memperbaiki incomplete trace. Algoritma yang memanfaatkan Model Markov Tersembunyi tersebut terdiri atas gabungan dari metode pembentukan model proses serta metode yang dimodifikasi. Metode yang dimodifikasi adalah metode Baum- Welch, Backward serta Viterbi, dimana pemodifikasian diperuntukkan untuk mengelompokkan observer pada Model Markov Tersembunyi dan menanggulangi aktivitas yang belum tercatat dalam Model Markov Tersembunyi. Uji coba dilakukan dengan menggunakan tiga periode process discovery, yaitu setiap 5 detik, setiap 10 detik dan setiap 15 detik. Hasil uji coba menunjukkan bahwa algoritma yang memanfaatkan Model Markov Tersembunyi mampu memperbaiki incomplete trace sehingga tidak ada aktivitas yang hilang dalam model proses. Hasil uji coba juga menunjukkan kualitas model proses dari algoritma yang memanfaatkan Model Markov Tersembunyi lebih baik dibandingkan kualitas model proses dari algoritma Online Heuristic Miner. ========== Process discovery is a technique for obtaining process model from sequences of activities recorded in the event log. Nowadays, information systems produce a streaming event log wherein it is recorded according to the execution time. Online Heuristic Miner is an algorithm of process discovery which capable of obtaining process models from the streaming event log. Disadvantage using an Online Heuristic Miner algorithm is the inability to overcome the incomplete trace on the streaming event log. The incomplete trace is a truncated trace, either clipped off at the beginning or clipped off at the end. The incomplete trace make a whole process cannot be displayed in the process model, so incomplete trace is noise in process discovery. An algorithm utilizing Hidden Markov Model is used to obtain a model process that can handle incomplete traces on a streaming event log. Hidden Markov Model is chosen because this model is widely used to predict the data which is useful in recovering incomplete trace. The algorithm utilizing Hidden Markov Model consists of a combination of methods of model process and modified methods. The modified methods are Baum- Welch, Backward and Viterbi methods, where the modification is intended to classify observer on Hidden Markov Model and combat activities that have not been recorded in the Hidden Markov Model. The experiments are using three period of discovery process, i.e. every 5 seconds, every 10 seconds and every 15 seconds. Experimental results show that the algorithm utilizing Hidden Markov Model is capable of recovering incomplete trace, so all activities are displayed in the process model. The experiment results also showed the quality of the process models obtaining by algorithm utilizing Hidden Markov Model is better than those obtaining algorithms Online Heuristic Miner.

Item Type: Thesis (Undergraduate)
Additional Information: RSIf 519.233 Sun p 3100016066340
Uncontrolled Keywords: Incomplete Trace, Kualitas Model Proses, Metode Baum-Welch, Metode Viterbi, Metode Backward, Model Markov Tersembunyi, Process discovery, Streaming event log, Backward Method, Baum-Welch Method, Hidden Markov Model, Incomplete Trace, Invisible Task, Process discovery, Streaming event log, Quality of Process Model, Viterbi Method.
Subjects: Q Science > QA Mathematics > QA274.7 Markov processes--Mathematical models.
Depositing User: - Davi Wah
Date Deposited: 29 Nov 2019 04:24
Last Modified: 29 Nov 2019 04:24
URI: http://repository.its.ac.id/id/eprint/71891

Actions (login required)

View Item View Item