Bee colony optimization for clustering incomplete data

Abstract

Many decision making processes include the situations when not all relevant data are available. The main issues one has to deal with when clustering incomplete data are the mechanism for filling in the missing values, the definition of a proper distance function and/or the selection of the most appropriate clustering method. It is very hard to find the method that can adequately estimate missing values in all required situations. Therefore, in the recent literature a new distance function, based on the propositional logic, that does not require to determine the values of missing data, is proposed. Exploiting this distance, we developed Bee Colony Optimization (BCO) approach for clustering incomplete data based on the p-median classification model. BCO is a population-based meta-heuristic inspired by the foraging habits of honey bees. It belongs to the class of Swarm intelligence (SI) methods. The improvement variant of BCO is implemented, the one that transforms complete solutions in order to improve their quality. The efficiency of the proposed approach is demonstrated by the comparison with some recent clustering methods.