Optimasi Penjadwalan Seminar Proposal Tugas Akhir Dengan Memperhatikan Kepakaran Dosen Penguji Berbasis Algoritma Modified Iterated Local Search Dan Tf/idf

Putra, Andaru Pratama Putra (2025) Optimasi Penjadwalan Seminar Proposal Tugas Akhir Dengan Memperhatikan Kepakaran Dosen Penguji Berbasis Algoritma Modified Iterated Local Search Dan Tf/idf. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026211187-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Penjadwalan atau timetabling problem dikenal sebagai sebuah masalah yang signifikan dalam bidang optimasi kombinatorik dengan kategori NP-hard. Dalam konteks universitas, penjadwalan melibatkan beberapa faktor antara lain ketersediaan ruangan, ketersediaan subjek, serta ketersediaan waktu. Pada Departemen Sistem Informasi (DSI) ITS, penjadwalan seminar proposal tugas akhir mahasiswa merupakan proses yang kompleks dan masih dilakukan secara manual. Hal ini berpotensi menyebabkan permasalahan seperti jadwal yang bertabrakan, proses penjadwalan yang memakan banyak waktu, serta keahlian dosen penguji yang tidak sesuai dengan topik tugas akhir mahasiswa. Tugas akhir ini bertujuan untuk mengembangkan sistem penjadwalan otomatis berbasis algoritma Modified Iterated Local Search (MILS) dan metode Term Frequency-Inverse Document Frequency (TF/IDF) untuk mempermudah proses penjadwalan seminar proposal tugas akhir mahasiswa DSI. Algoritma MILS dimodifikasi melalui penerapan strategi threshold adaptif dalam menerima sebuah solusi, sehingga akan meningkatkan eksplorasi ruang solusi dan menghindari local optima. TF/IDF digunakan untuk mencocokkan topik dari tugas akhir mahasiswa dengan kepakaran dosen penguji sehingga mampu meningkatkan kualitas dari jadwal yang dihasilkan. Hasil uji coba menunjukkan bahwa sistem yang diusulkan berhasil menghasilkan jadwal seminar proposal feasible atau memenuhi semua hard constraint yang ditentukan. Perbandingan hasil uji coba MILS dan ILS menghasilkan penurunan 23% dari penalti akhir MILS yaitu 62,09 terhadap penalti akhir ILS yang bernilai 80,11 pada skenario 1000 iterasi. Algoritma MILS terbukti lebih unggul secara signifikan dibanding algoritma Iterated Local Search (ILS) standar, dengan penurunan penalti sebesar 40% dari penalti solusi awal MILS yang bernilai 94,51 pada uji coba 1000 iterasi yang lain. Analisis sensitivitas parameter variabel ambang batas menunjukkan bahwa strategi threshold adaptif mampu menghasilkan solusi dengan kualitas terbaik dibandingkan dengan penggunaan threshold statis.
==================================================================================================================================
Scheduling, or timetabling problem, is recognized as a significant problem in combinatorial optimization, classified as NP-hard. In the university context, scheduling involves various factors such as room availability, subject availability, and time constraints. At the Department of Information Systems (DSI) ITS, scheduling proposal seminars for undergraduate theses is a complex process and still conducted manually. This manual approach often leads to issues such as scheduling conflicts, time-consuming processes, and mismatches between examiner's expertise and student's thesis topics. This research aims to develop an automated scheduling system using the Modified Iterated Local Search (MILS) algorithm and the Term Frequency-Inverse Document Frequency (TF/IDF) method to streamline the scheduling process for undergraduate thesis proposal seminars at DSI. The MILS algorithm is enhanced by applying an adaptive threshold acceptance strategy, improving solution space exploration and avoiding local optima. TF/IDF is used to match thesis topics with the expertise of examiners, thereby enhancing the quality of the generated schedule. The expected results of this algorithm's implementation in this research is a feasible seminar schedule that fulfills all hard constraints and optimizes as many soft constraints as possible. Experimental results show that the proposed system successfully produces feasible seminar schedules that meet all the defined hard constraints. Comparative testing between MILS and standard ILS resulted in a 28% reduction in final penalty, with MILS achieving a penalty of 59.05 compared to ILS’s 80.11 with 1000 iterations. Furthermore, MILS achieved a 40% improvement from its initial solution with a starting penalty of 94.51 in a different 1000 iterations experiment. Sensitivity analysis of the threshold parameter also indicates that the adaptive threshold strategy consistently generates higher-quality solutions than static threshold usage. This system is anticipated to support academic processes at DSI ITS by improving both the efficiency and quality of undergraduate thesis proposal seminar scheduling.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Penjadwalan, Timetabling problem, Modified Iterated Local Search, Term Frequency-Inverse Document Frequency, Seminar Proposal, Scheduling, Timetabling Problem, Modified Iterated Local Search, Term Frequency-Inverse Document Frequency, Proposal Seminar
Subjects:	T Technology > T Technology (General) > T57.84 Heuristic algorithms.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Andaru Pratama Putra
Date Deposited:	25 Jul 2025 02:45
Last Modified:	25 Jul 2025 02:45
URI:	http://repository.its.ac.id/id/eprint/121070

Actions (login required)

View Item