K-medoids Clustering of Data Sequences with Composite Distributions