Clustering by Sorting Potential Values (CSPV): A novel potential-based clustering method

A novel clustering method called Clustering by Sorting Potential Values (CSPV) is proposed. The clustering is done in an efficient tree-growing fashion based on both the distances and the hypothetical potential values produced from the distribution of all the data points. The method is simple but is shown to be very effective in identifying different kinds of clusters. It outperforms four popular clustering methods in most of our experiments and is the only one that works for all the six studied data sets. Moreover, it is designed as a generic method which can be easily applied to different clustering problems.

► We presented an entirely new clustering method which can find the number of clusters automatically.
► A hypothetical potential field is used to capture the effects of both the global and local data distributions on a data point.
► Order of the data points given by their potential values is found to be helpful in the clustering process.
► New algorithm was able to outperform four popular clustering methods in most our experiments.