CLUSTERING CATEGORICAL DATA USING k-MODES BASED ON CUCKOO SEARCH OPTIMIZATION ALGORITHM

TLDR: This paper proposes a new k-modes clustering algorithm that is combined with Cuckoo Search algorithm to obtain the global optimum solution for categorical clustering and shows the efficiency of the proposed algorithm.

Abstract: Cluster analysis is the unsupervised learning technique that finds the interesting patterns in the data objects without knowing class labels. Most of the real world dataset consists of categorical data. For example, social media analysis may have the categorical data like the gender as male or female. The k-modes clustering algorithm is the most widely used to group the categorical data, because it is easy to implement and efficient to handle the large amount of data. However, due to its random selection of initial centroids, it provides the local optimum solution. There are number of optimization algorithms are developed to obtain global optimum solution. Cuckoo Search algorithm is the population based metaheuristic optimization algorithms to provide the global optimum solution. Methods: In this paper, k-modes clustering algorithm is combined with Cuckoo Search algorithm to obtain the global optimum solution. Results: Experiments are conducted with benchmark datasets and the results are compared with k-modes and Particle Swarm Optimization with k-modes to prove the efficiency of the proposed algorithm.