Experimental Code for KDD 2017 For paper: Scalable Top-N Local Outlier Detection If you find any problems, please contact yyan2@wpi.edu.
Includes three parts:
-
Baseline Methods
Two different indexing methods for KNN search are implemented.
--- Pivot-based KNN Search :
Main Class: baseline.pivotknn.ComputeTopNLOF From Paper: Bhaduri, Kanishka, Bryan L. Matthews, and Chris R. Giannella. "Algorithms for speeding up distance-based outlier detection." Proceedings of the 17th ACM SIGKDD international conference on Knowledge Discovery and Data Mining. ACM, 2011.
--- R-Tree based KNN Search:
Main Class: baseline.rtreeknn.ComputeTopNLOF From Paper: Guttman, Antonin. R-trees: a dynamic index structure for spatial searching. Vol. 14. No. 2. ACM, 1984.
-
MicroCluster --- State-of-the-art method:
Main Class: microcluster.topnlof.TopNLOFDetection From Paper: Jin, Wen, Anthony KH Tung, and Jiawei Han. "Mining top-n local outliers in large databases." Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 2001.
-
TOLF --- Proposed method
Two versions of TOLF: Single-thread and Multi-thread version. Main Class for Single-Thread Version: cellpruning.lof.pruning.ComputeTopNLOFWithPruning Main Class for Multi-Thread Version: cellpruning.lof.pruning.MultiThread.ComputeTopNLOFWithPruning From Paper: Scalable Top-N Local Outlier Detection (In Submission)