Genetic Algorithm-based Classification Rule Mining

Data Mining is a new research field rising in recent years, and involves an integration of techniques such as database and warehouse technology, statistics, machine learning, artificial intelligence etc. Its goal is to discover valuable information and knowledge hidden behind data from numerous data in order for providing decision support. Mining classification rules is a procedure to construct a classifier through studying training dataset, and is a very important part of Data Mining and Knowledge Discovery. In essence, its goal is to discover classification rules which are highly predictive accurate, comprehensible and interesting.In this paper, we give an introduction to the basic theories of data mining and genetic algorithm. Then, we lay emphasis on studying the application of genetic algorithm in classification rule mining. In order to overcome the premature phenomena, based on the SGA, this thesis introduces the idea of " non-random initial population " and " uniform operator ", and puts forwards Mining Classification Rules Based on Genetic Algorithms with Non-random Initial Population, and applies the new algorithm into mining classification rules on Breast cancer data and Dermatology Data. At the same time, this thesis improves simple genetic algorithm by multi-objective genetic algorithm based on realistic demand, proposes classification rule mining based on multi-objective genetic algorithm and tests the algorithm on adult database and zoo database. From the experimental results, it was observed that, these methods can guarantee to get rid of any local solution when handled the problems of GAs in the task of classification. It can also mostly improve the comprehensibility of the discovered knowledge.