Use Data Mining to Improve Genetic Algorithm Efficiency for a Job Shop Scheduling Problem

This paper proposes a new improved Genetic Algorithm (GA) by utilizing a Data Mining technique, and demonstrates how it is superior to traditional GA on a popular job shop scheduling problem. GA has long been widely applied to solve complex optimization problems in a good variety of areas. It has advantages of adaptive capability, efficient search, potential to avoid local optimum, etc. In recent literature, researchers have proposed a good number of new GAs by combining basic GA with other techniques, such as heuristic rules, simulated annealing, neural networks, fuzzy sets, and so on, in order to improve the efficiency for various optimization problems. Data mining is a new evolving technology for knowledge extraction, classification, clustering, estimation, etc. The capability of finding frequent patterns in large data set is the key reason why it is integrated with GA in this research. Due to the fundamental concept of GA’s randomness during evolution, a traditional GA may become less efficient in search for optimum. By embedding the frequent schemata into the GA evolution process, the new improved GA could reduce the search time by preserving segments of good solutions without accidentally being lost due to random crossover or mutation. The proposed new GA was experimented on a popular 6x6 job shop scheduling problem. The results have shown its better efficiency than traditional GAs and potential for further research works.