Research Outlier Detection Technique Based on Clustering Algorithm

In this paper, in clustering and outlier detection as a starting point, it put forward a kind of DBSCAN-LOF algorithm, to the core definition of object DBSCAN, then the LOF only need to operate on noncore object, thereby reducing the number of the original LOF algorithm for global object operation, the results show that the algorithm improve the running efficiency of the LOF, and the clustering effect of DBSCAN, and the at the same time, the clustering and outlier detection results is produced.