Robust Automatic Clustering Based on Local Density with Glowworm Swarm Optimization

Robust clustering based on nature inspired algorithms have gained a lot of attention due to their ability to effectively find solutions for data clustering problems while detecting and eliminating outliers. This paper proposes a robust automatic clustering algorithm based on Glowworm Swarm Optimization (GSO). The GSO is a meta-heuristic algorithm inspired from the ability of glowworms to change their luminescence and thus to glow at different intensities based on their capability (which represents fitness value). The proposed algorithm, termed as Robust Glowworm Swarm Clustering (RGSC), is capable of capturing multiple local maxima of the fitness function, which are representing cluster centers. The proposed algorithm has been assessed on six artificial as well as eight real datasets having number of clusters varying from 2 to 10. The superior performance is reported in terms of evaluation metrics like Silhouette Score and Adjusted Rand Index, as compared to existing clustering algorithms K-means, Mean-shift, DBSCAN, and Robust PSO.