In many practical applications, it is very important to select a good feature extraction method (not necessarily the best clustering algorithm) that highlights the underlying structure of the dataset from clustering aspect.
•
Many practical datasets contain patterns which are similar and overlapping in nature. In such cases the transformed domain information of the dataset can be effectively used to group the patterns.
•
In many cluster analysis applications there is a need for stability or consistence performance of the results. As most nature-inspired algorithms are heuristic in nature, stability issues of these clustering algorithms are still a barren area of research.