Geospatial outlier detection method based on multivariate adaptive regression

Abstract

The invention discloses a geospatial outlier detection method based on multivariate adaptive regression which gives fully consideration on correlativity and heterogeneity of geospatial data. Data preprocessing and regression variables determining are performed according to acquired detection variables and other geographic variables to construct an adaptive spatial neighborhood, a geographically weighted regression model is constructed through weight function selection based on adaptive bandwidth and by repairing thematic attribute variation gradients of spatial neighboring entities, robust spatial outlier measures of all the spatial entities are calculated and combined in a set, and one sampled point that has an outlier certain times deviated from an average is determined as a spatial outlier. The method of the invention has the advantages that adaptive bandwidth selection for spatial correlativity and quantitative expression of the relationship between detection variables for spatialheterogeneity and other geographic variables are taken into consideration and the method is more practical and explanatory.

Description

technical field [0001] The invention belongs to the field of spatial data mining and spatial analysis, in particular to a method for detecting anomalies in geographic space based on multiple self-adaptive regression Background technique [0002] Anomaly detection originally originated from the research of gross error detection in statistics, but in practice it is found that some "gross errors" are not necessarily errors, but often imply some special laws or properties, which have important application value. Spatial anomaly detection is the expansion and extension of traditional anomaly detection methods in the field of geoinformation science. As an important means of spatial data mining, it aims to find a small part of anomalous entities that deviate from the overall or local general distribution pattern from massive spatial data. Abnormal entities usually contain special development laws of geographical phenomena or geographical processes. Hawkins was the first to give th...

AI Technical Summary

Problems solved by technology

Existing studies have shown that the choice of weight function has little effect on geographically weighted regression models and is very sensitive to the choice of bandwidth

And when the data points participating in the regression contain potential abnormalities, it will have a greater impact on the regression results

At the same time, most of the existing multivariate spatial anomaly detection does not take into account the differences in the contribution weights of different influencing variables, and does not consider the impact of spatial heterogeneity on regression factors

[0005] In the process of realizing the present invention, the inventors have found that there are at least the following problems in the prior art: in geospatial anomaly detection, it is difficult to take into account the robust anomaly measurement of multivariate data, and it is difficult to deal with potential anomaly effects and adaptive selection in geographically weighted regression models bandwidth

Images