A learning method of scheduling knowledge by genetic algorithm

Proposes a learning method of the status selection knowledge using GAs (genetic algorithms). The GA has been traditionally used for getting the solution in past researches on scheduling. However, the application of GA to getting the solution causes reproduction problems, in which the solution generated from the same scheduling problem is not always the same. In addition, a great deal of time is needed to solve each problem. On the other hand, in the authors´ proposed method, the GA is applied to learning scheduling knowledge. Though it takes a great deal of time in learning the knowledge, little time is needed to solve problems in the daily operation phase, and the reappearance feature is guaranteed. The representation of scheduling knowledge by gene is usually difficult, because it is necessary to carry out crossover operation on a gene and the representation of knowledge is much more complex than the representation of the solution of a planning problem. To cope with the representation problem, the authors introduce tree construction for representing status selection knowledge