关键词: 共享最近邻聚类算法, 一趟聚类算法, 大规模数据集

Abstract: Clustering is a method of unsupervised learning in machine learning，the typical task of which is to discovery “natural” clusters present in the data.The shared nearest neighbor algorithm is one of the most efficient clustering algorithm which can handle datasets of different sizes，shapes and densities.But there are still some shortages about the algorithm.SNN can’t handle large dataset because of its high complexity.There are no definite methods about threshold of the algorithm.SNN can not process databases with mixture attributes.This paper improves the SNN algorithm to process the data with categorical attributes，gives a simple definite method to select threshold of the algorithm.The time complexity of the improved algorithm is nearly linear with the size of dataset and can be used to large dataset.The experimental results on real datasets and synthetic datasets show that the improved algorithm is effective and practicable.

Key words: shared nearest neighbor clustering algorithm, one-pass clustering algorithm, large dataset