Metric space outlier detection method based on various supporting points

Abstract

The invention discloses a metric space outlier detection method based on various supporting points. The method comprises the steps of selecting a distance function, the distance function is selected according to the data type of a data set; the step of selecting the supporting points, wherein the data set is read, and the non-repeated intensive supporting points and the edge supporting points are selected; the step of establishing an index, wherein the distances between all objects and the dense supporting points are calculated and sequenced to form the one-dimensional index, the distances between all the objects and the edge supporting points are calculated and marked as second distances, first distances and the second distances serve as coordinates, and a supporting point space is formed; the step of conducting outlier detection, wherein the one-dimensional index is divided into a plurality of data blocks, and outlier detection is performed on the data blocks one by one. Due to the fact that the dense supporting points and the edge supporting points are selected at the same time, the technical problem that the data space is distorted due to a single supporting point is avoided, meanwhile, the time and cost for establishing the index are reduced, the distance calculation frequency is reduced, and the outlier detection speed is increased. The method is used for detecting outlier points in the data set.

Description

technical field [0001] The invention relates to the field of data mining, in particular to a metric space outlier detection method based on multiple support points. Background technique [0002] Outliers are unusual data points in a data set that behave so differently from other points that one suspects that these data are not random deviations, but are produced by an entirely different mechanism. Outliers are also called abnormal points or abnormal objects. Outlier detection is also called anomaly detection, deviation detection or outlier mining, which is to detect outliers in the data set according to a certain algorithm. In other words, outlier detection is to mine the very few points in the massive data that are significantly different from the mainstream data. [0003] Most of the traditional outlier detection techniques are oriented to multi-dimensional space, which are only applicable to multi-dimensional data, and are helpless for complex data types such as images,...

AI Technical Summary

Problems solved by technology

[0007] The above-mentioned HIOD algorithm overcomes the shortcomings of the iORCA algorithm, but this algorithm only selects one support point to achieve the goal of dense support points and edge support points at the same time, and it takes a long time to build an index

Images