In order to solve the defect in the spatial outliermining algorithm that the spatial objects may be affected bytheir surrounding abnormal neighbors, a Based K-NearestNeighbor (BKNN) algorithm was proposed based on theworking principle of KNN Graph, which could effectivelyidentify the spatial outliers by using cutting edge strategies.The core idea of BKNN is to calculate the dissimilarity ofthe non-space attribute values the between adjacent objects,and to find the find the largest local outlier or outlierregions by cropping off the edges with the largestdissimilarity. The experiments for the spatial outlier miningalgorithm BKNN based on the KNN Graph were carried outin the real datasets FMR and WNV. The example of thealgorithm and the time complexity were analyzed and theresults were compared to those of the existing classicalalgorithms, which verified that this algorithm could improvethe accuracy of spatial outlier mining and simultaneouslymine spatial region outliers.
展开▼