| Autor: |
Muhammad Fachrie, Anita Fira Waluyo |
| Jazyk: |
angličtina |
| Rok vydání: |
2021 |
| Předmět: |
|
| Zdroj: |
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi), Vol 5, Iss 3, Pp 550-556 (2021) |
| Druh dokumentu: |
article |
| ISSN: |
2580-0760 |
| DOI: |
10.29207/resti.v5i3.2988 |
| Popis: |
One of the many techniques used to solve the University Course Timetable Problem (UCTP) is Genetic Algorithm (GA) which is a technique in the field of Evolutionary Computation. However, GA has high computational complexity due to the large number of evolutionary operators that must be performed during the evolutionary process, so it takes a long time to produce an optimal timetable. The computation time will also increase when the number of optimized variables is very large, such as in UCTP. Of course, this makes the application less reliable by users. Therefore, this article proposes a parallelization model for GA to reduce computation time in solving UCTP problems. The proposed AG is designed with a multithreading CPU scheme and implements a guided creep mutation mechanism and eliminates the recombination mechanism to reduce more computation time. The proposed system was tested and evaluated using two different UCTP datasets from the University of Technology Yogyakarta which contained 878 and 1140 lecture meetings in even and odd semesters. Unlike the previous ones, this study discusses UCTP with dynamic time slots where the duration of the lecture depends on the course credits. From the tests that have been done, it is found that the GA that was built is able to generate optimal course timetable without any clashes in a relatively fast time, that is less than 60 minutes for 1140 lecture meetings and less than 20 minutes for 878 lecture meetings. The use of the multithreading CPU model has succeeded in reducing computation time by 62% when compared to the conventional model which only uses one thread. |
| Databáze: |
Directory of Open Access Journals |
| Externí odkaz: |
|
