\input zb-basic
\input zb-ioport
\iteman{io-port 06041219}
\itemau{Chen, Gang; Cai, Yuanli; Mu, Jing; Yang, Weili}
\itemti{An ordinal anomaly probability algorithm for anomaly detection problems in massive data sets.}
\itemso{J. Xi'an Jiaotong Univ. 45, No. 4, 36-40 (2011).}
\itemab
Summary: An ordinal anomaly probability (OAP) method is proposed to improve the efficiency, effectiveness and stability of existing anomaly detection algorithms. Since anomalies are easier to isolate by uniformly partitioning the data space, the order of anomaly probabilities can be evaluated in terms of isolation depths in uniform $N$-ary partition trees. Then, the $k$ largest probable anomalies are extracted. The OAP method can ignore the evaluation of distance and density, and hence reduces the complexity to $O(n)$. Experimental results show that the CPU time of the OAP method increases linearly with a linearly growing data set. Furthermore, comparisons show that the OAP method is 30 times faster than iForest and is much more stable, while its accuracy is improved by 20\%--30\%.
\itemrv{~}
\itemcc{}
\itemut{data mining; anomaly detection; uniform partition; ordinal anomaly probability}
\itemli{}
\end