Abstract: When the k-nearest neighbor method is used,it is difficult to choose an appropriate parameter k of the algorithm which affects obviously its efficiency and performance.Natural nearest neighbor proposed by us is a novel concept on nearest neighbor,in which each point’s neighbors are formed by an adaptive algorithm without parameter k.In this paper,we proposed an outlier detection algorithm based on natural nearest neighbor(ODb3N) by means of modifying iteration stop condition．The experiments show that our method not only has the advantage of non-parameter,but also has the ability to discover both the outlier and the cluster of outliers.

Key words: k-nearest neighbor,Natural nearest neighbor,Outlier detection,Cluster of outliers