Research On Anomaly Detection Method Based On DBSCAN Algorithm

Anomaly detection is a commonly used method in the field of data mining,and cluster-based anomaly detection technology is the most popular.A variety of clustering algorithms have been proposed,which are mainly divided into partitioning,hierarchical and density-based clustering algorithms.Density-based clustering can process cluster structures of arbitrary shapes and is easy to implement,so the algorithm becomes very popular,and the DBSCAN algorithm is the most representative.It is also often used for the detection of anomalous data due to its performance in identifying noise data.However,in practical applications,the DBSCAN algorithm has higher time complexity and lower execution efficiency,which is not conducive to processing massive data sets;and it is sensitive to input global parameters,which is not conducive to processing non-uniform data sets.Based on the above problems,this paper proposes a solution.The main work includes the following aspects:Firstly,in order to reduce the running time of the algorithm,combined with the MCMC(Markov Chain Monte Carlo)sampling method,DBSCAN(Density-based spatial clustering of applications with noise,DBSCAN)is improved.A new clustering algorithm called DBSCAN++.The basic idea is to prioritize the expansion of core objects with strong expansion capabilities.Theoretical analysis and simulation experiments show that DBSCAN++ has comparable accuracy with respect to DBSCAN clustering,while DBSCAN++ has lower runtime.Therefore,DBSCAN++ is an effective clustering algorithm.Secondly,according to the DBSCAN++ parameter sensitivity problem,combined with the kernel density estimation theory,an improved algorithm based on adaptive strategy DBSCAN++ is proposed,which is called A-DBSCAN++.The basic idea is to dynamically determine the local density threshold MinPts of the extended node based on the kernel density estimation theory.In the experiment,by comparing with DBSCAN++,the results show that the accuracy of A-DBSCAN++ is improved,and the running time is still kept.