2015 International Symposium of Computer Application and Information Technology (IsCait 2015)

251 / 1971-01-01 00:00:00

A Mixed Numerical And Categorical Data Clustering Method Combining Genetic Algorithm And Partition Similarity

5642

In real world, data objects described by both numerical and categorical attributes encountered commonly. K-prototype (KP) is one of the most effective algorithms for clustering this kind of data. However, it highly depends on the initial value selection and converges to local optimum easily. In this paper, a Genetic Algorithm based K-prototype method (GAKP) is introduced, in which KP is applied for local searching under the framework of Genetic Algorithm. And a new partition similarity based fitness function is employed. Experiments on benchmark datasets show that the proposed method can get much better results and is more robust than that of traditional clustering algorithms.