Improvement Of Density-Based Local Outlier Detection Algorithm

In recent years,science and technology have developed faster and faster,and information dissemination has become more widespread.The technology of data mining has been promoted in various aspects,and people are also greatly benefited here.The purpose of data mining is to extract knowledge that is previously unknown but potentially valuable in a massive,ambiguous data set.As a hot branch of data mining,outlier detection is mainly used to find objects that deviate significantly from or do not satisfy the behavior characteristics of general objects.At present,outlier detection is widely used in many fields.Traditional outlier detection algorithm can detect abnormal data objects which are suitable for its own algorithm to a certain extent.However,there are some shortcomings and defects,such as low detection efficiency,low detection accuracy.In this paper,the traditional density-based outlier detection algorithm has higher computational time complexity and is only applicable to datasets of a certain scale.When large-scale datasets are encountered,the precision is usually low,and the calculation is repeated during the mining process.Excessive steps and other issues,and proposed an outlier detection method based on square neighborhood and cropping factor.Firstly,square neighborhood is used to absorb the idea of grid algorithm,and expanded square neighborhood is used to replace grid segmentation.Clustering points are eliminated quickly,thus avoiding the "disaster-sustaining" problem of grid algorithm.Secondly,in order to improve the accuracy of the algorithm,the concept of clipping factor is introduced to select candidate outliers.Then,a new measure is given to evaluate the degree of outlier.In order to verify the effectiveness of the improved algorithm,simulation experiments are carried out on real data sets and synthetic data sets.Different outlier detection algorithms are compared and analyzed.The experimental results show that the improved algorithm can effectively identify outliers and has obvious advantages in execution efficiency.Finally,the improved algorithm is applied to network intrusion detection.The properties,attack types and distribution of KDDCUP99 data are analyzed,and the intrusion data is preprocessed.The experiment shows that the algorithm has a good application prospect in intrusion detection.