Two Medoid-Based Algorithms for Clustering Sets

This paper proposes two algorithms for clustering data, which are variable-sized sets of elementary items. An example such data occurs in the analysis a medical diagnosis, where goal is to detect human subjects who share common diseases possibly predict future illnesses from previous history. The first proposed algorithm based on K-medoids and second extends random swap algorithm, has proven be capable efficient careful clustering; both depend distance function among objects (sets), can use application-sensitive weights or priorities. makes it possible exploit several seeding methods that improve accuracy. A key factor their parallel implementation Java, functional programming using streams lambda expressions. parallelism smooths out O(N2) computational cost behind indexes as Silhouette index allows handling non-trivial datasets. applies benchmark case studies demonstrates how accurate time-efficient solutions achieved.