D. Pumphrey and M. Lazarescu (Australia)
Machine Learning, Data Mining, Knowledge Acquisition.
In this paper we describe a clustering algorithm that uses exemplar points to represent the clusters discovered in the data. The algorithm uses incremental learning to select and update the set of exemplar points. The clustering of the data is based on the CHAMELEON algorithm which uses a k-sparse graph approach. Our algorithm has the advan tages that it can handle clusters that have irregular shapes and varying sizes while still allowing an incremental pro cessing of the data. We describe the algorithm and present results obtained from generated datasets. We also compare its clustering performance with the CURE algorithm which is another algorithm that uses representative points to clus ter data.
Important Links:
Go Back
