Mining knowledge in large scale databases using cultural algorithms with constraint handling mechanisms

This paper proposes a framework for evolutionary systems to mine implicit knowledge in large scale databases. The idea here is to construct knowledge-based evolutionary systems that apply the power of evolution computation to facilitate the data mining processes. This framework provides the possibility of making two processes, the data mining process and the optimization process, work simultaneously and reciprocally. Based on Cultural Algorithms, the data mining process is supported by symbolic reasoning in the belief space, and the optimization process is supported by evolutionary search in the population space. The evolutionary search in databases can facilitate the data mining process, while the data mining process can also provide knowledge to expedite the search in databases i.e. the data mining process and the evolutionary search can be integrated and benefit from each other. This new approach was applied to a large-scale temporal-spatial database, and the results indicate that it successfully mined out some very interesting patterns that are unknown before. Another advantage of this approach is that it doesn't have to access all information in the database in order to identify some interesting patterns, by automatically "select" useful cases from a large database to avoid the exhaustive search to every cases. This suggests a great potential to reach the goal of efficiency and effectiveness for data mining.