Measuring similarity between sequential datasets.

Similarity measurement is a basic problem in data mining, but little work focuses on the similarity between sequential datasets. We propose the density-emerging pattern. And we propose a novel similarity measurement between sequential datasets based on the quality of shared-density-aware and shared-emerging patterns. Similarity measuring can be divided into three stages, i.e., pattern mining, evaluating the quality of patterns, and evaluating similarity. We performed experiments on real protein sequence datasets to test the effectiveness and efficiency of our method. A case study of sequential data set classification was carried out and high accuracy was obtained. The results show that our method is able to be effectively used in the classification of sequential datasets.