E-Transitive: an enhanced version of the Transitive heuristic for clustering categorical data
Said Chah Slaoui, Zineb Dafir, Yasmine Lamari · Procedia Computer Science · 2018
Clustering is among the most widely used data mining tasks since it plays an important role in organizing huge amounts of data. It aims to form natural groupings by partitioning sets of data objects into disjoint and homogeneous groups. The present paper proposes an enhanced version of a recently appeared heuristic for clustering large categorical datasets, called Transitive heuristic. The improvements brought to the original heuristic concern mainly the calculation of the representatives of clusters and the manner in which each data object is processed. Experiments based on real-life datasets demonstrate that the proposed version yields more accurate results.