SLPPCA: A New Efficient Cluster Algorithm Based on SLPP

Sometimes it is quite difficult for traditional clustering algorithms to find useful clusters in large data base (DB), because that each cluster may be different with each other on their shape, size, and density and so on. Here the paper provides a novel efficient algorithm named SLPPCA (stem-leaf-point plot clustering algorithm) which is a good solution to those problems. The main idea of SLPPCA is to regards each cluster of data objects as a closed area with its inner objects enclosed with its boundary points the algorithms identified out from all objects and grouped to each cluster automatically by SLPP. The SLPP can provide us directly with congregational, dispensational and distributional characteristics of DB, and with the help of those traits and the method of exploratory data analysis (EDA) , we can divide data objects into groups (clusters) just rely on data itself and without defining any transcendental parameters such as the cluster number or similarity threshold value which are needed for most of the traditional cluster algorithms, the yielding groups maybe in unlike shapes, different size or disparate density. The result shows that SLPP is excellent, which can adapt well for arbitrary shapes, size and density. Moreover, the SLPPCA can be programmed in parallel well and run in even high efficiency.