An Inflection Point Based Clustering Method for Sequence Data.

With rapid development of information technology, large amount of sequential data has been accumulated. How to extract business-related knowledge from sequential data set has become an urgent problem. The data pattern of sequential data is similar to that of traditional relation data, but there is temporal association between different attributes. It makes the traditional distance measurement method fails, which leads to the difficulty of existing clustering methods to apply. Aiming at above problems, the concept of data inflection point is introduced, and a method of measuring the dissimilarity of inflection points is proposed to realize inflection points marking, leveraging density-based clustering. Further, data sequence clustering method, called DSCluster, is proposed by applying frequent item set mining algorithm Apriori to the marked inflection point sequence set. Theoretical analysis and experimental results show that DSCluster method can effectively solve the problem of sequence data clustering. It can fully take into account the sequence of data related features extracting, the algorithm is effective and feasible.