¹Hebei Normal University of Science & Technology, China
²Foreign Language College of Dalian Jiaotong University, China
³College of Information Science and Engineering Yanshan University, China

---

Abstract—In order to solve the defect in the spatial outlier mining algorithm that the spatial objects may be affected by their surrounding abnormal neighbors, a Based K-Nearest Neighbor (BKNN) algorithm was proposed based on the working principle of KNN Graph, which could effectively identify the spatial outliers by using cutting edge strategies. The core idea of BKNN is to calculate the dissimilarity of the non-space attribute values the between adjacent objects, and to find the find the largest local outlier or outlier regions by cropping off the edges with the largest dissimilarity. The experiments for the spatial outlier mining algorithm BKNN based on the KNN Graph were carried out in the real datasets FMR and WNV. The example of the algorithm and the time complexity were analyzed and the results were compared to those of the existing classical algorithms, which verified that this algorithm could improve the accuracy of spatial outlier mining and simultaneously mine spatial region outliers.

Index Terms—Spatial outlier; Spatial region outliers; KNN Graph; BKNN algorithm.

[PDF]

Cite: Lijun Cao, Xiyin Liu, ZhiPing Wang, Zhongping Zhang, "The Spatial Outlier Mining Algorithm based on the KNN Graph," Journal of Software vol. 8, no. 12, pp. 3158-3165, 2013.