Automatic shape independent clustering inspired by ant dynamics

This article describes a simple heuristic algorithm that can automatically detect any number of well-separated clusters, which may be of any shape e.g. convex and/or non-convex. This is in contrast to most of the existing clustering algorithms that assume a value for the number of clusters and/or a particular cluster structure. The algorithm draws inspiration from the dynamics of ants and iteratively partitions the dataset based on its proximity matrix. A runtime complexity analysis shows that the complexity of the algorithm is either quadratic or cubic with respect to the size of the dataset. It can detect outliers from the data and is also able to identify the situation when the data do not have any natural clusters at all. Promising results on both real and artificial datasets have been included to show the effectiveness of the proposed technique.