Information Systems International Conference (ISICO), 2 – 4 December 2013 Copyright © 2013 ISICO Solving University Timetabling Problem Using Harmony Search: a Case Study in STMIK Mikroskil Harrison, Hardy, Ng Poi Wong Harrison, Hardy, Ng Poi Wong Department of Computer Science, STMIK Mikroskil Keywords: Harmony Search University Timetabling Machine Learning Optimization ABSTRACT A University Timetabling Problem (UTP) belongs to NP-complete (Nondeterministic Polynomial-Complete) problem category since it has huge search space that consists of many students, classrooms, lectures, and courses permutation. Due to that, finding an optimal schedule using exact algorithm is impossible to be done under a reasonable time. Harmony Search is a heuristic method that can solve NP-Complete problem such as UTP in reasonable time. However because of its heuristic nature, it can only provide solution which is close to the perfect solution. Therefore, to prove that a schedule is a feasible solution, several constraints have to be satisfied. This research used Harmony Search to solve the UTP whereby the dataset, constraints, and experiments are acquired and performed in STMIK Mikroskil. This research also evaluates the speed performance and fitness value of Harmony Search using different parameter configuration. From its finding, the HMCR parameter value that is best suited for scheduling is in range of 0.7 to 0.95. The PAR value however, did not sufficiently contribute to the speed and fitness value. Copyright © 2013 Information Systems International Conference. All rights reserved. Corresponding Author: Harrison, Department of Computer Science, STMIK Mikroskil, Jalan Thamrin No. 140, Medan, Indonesia. Email: conan_teleport@hotmail.com 1. INTRODUCTION A University Timetabling Problem (UTP) is a placement of a set of courses in predefined time period [1]. UTP consists of lecture and exam scheduling where both of them face common problem of finding a schedule over a limited time period while avoiding conflicts and satisfying a number of side- constraint [2]. UTP has long been known as NP-complete problem class [3]. NP-complete problem class can’t be solved using exact algorithm since it has a huge search space whereby iterating each possible solutions can’t be done under a reasonable time period. However, heuristic algorithms such as Genetic Algorithm [4], Particle Swarm Optimization [5], Harmony Search (HS) [6] and many others can be used to solve NP-complete problem. These heuristic algorithms can provide an optimal solution to UTP. HS is chosen as the preferred method of this research due to its performance that is better than other heuristic algorithms [6]. This research is perfomed in STMIK Mikroskil environment. All the data that are used as input to HS are collected from STMIK Mikroskil. The resulting schedule also have to conform to the constraints that are stated in the STMIK Mikroskil academic rules. The contribution from this research is majorly directed toward the academic improvement of STMIK Mikroskil and the evaluation of speed performance and fitness value of HS using different parameter configurations that are best suited for scheduling purpose. 2. LITERATURE REVIEW University Timetabling Problem has been investigated in several researches such as Al-Betar et al. [7], Nguyen et al. [8], and Chen and Shih [9]. In Al-Betar et al. [7]’s work, HS is used with multi-pitch adjusting rate. Nguyen et al [8] used a hybrid algorithm of HS and Bees [10] algorithm. Chen and Shih [9]