Abstract
A timetable must be made at least once, and probably several times, every year in all kinds of educational institutions including elementary, junior high and senior high schools, and universities. At the junior and senior high school levels, many teachers only teach one or more subjects for some classes, and thus adjustments of some classes and lessons need to be taken into consideration. This problem is a kind of assignment problem which we refer to as the school timetabling problem. In a junior high school that has 24 classes, it takes about one week for seven planners to complete the school timetable, and this work is performed twice a year. In a past study, we applied the tabu search to the timetabling problem for junior high school, and proposed three algorithms. In addition, we showed that the weight for the violation of constraints greatly influenced the search. However, one problem was that it was very difficult for planners to find the best weight setting. In this paper, we propose some methods for automatically adjusting the weight setting during tabu search. When the solution is not improved over a period, the weight setting is changed according to each rule. By applying these methods to a numerical experiment for an existing junior high school, we show that they impove search efficiency. In particular, the method that depended on the number of violation of constraints when changing the weight setting is more effective. This indicates that planners do not need to be conscious of the weight setting. In addition, the timing of weight change and application of the algorithm in a past study are described.
