213 多目的遺伝的アルゴリズムを用いた多目的クラスタリング

抄録

Data mining is the process to find some patterns or rules from the large volumes of data automatically. It can be classified into six fields ; classification, estimation, prediction, association rule, clustering and profiling. We focus on the data clustering and classify the similar data into the same clusters. These clusters can be used in the field of preparation of data analysis, finding segmentations in market and so on. To derive the clusters, we apply "Multiobjective clustering with automatic determination of the number of clusters (k): MOCK". MOCK uses two complementary objectives based on cluster compactness and connectedness, and returns a set of different trade-off partitioning over a range of different cluster numbers, k. It is able to find the appropriate number of clusters based on the information of the trade-off curve. In this paper, we proposed the scalable k auto-determination scheme of the number of clusters. The proposed scheme reduced the Pareto-size and usually select adjust number of clusters.