摘要
A novel approach for outlier detection with iterative clustering( ICOD) in diverse subspaces is proposed. The proposed methodology comprises two phases,iterative clustering and outlier factor computation. During the clustering phase, multiple clusterings are detected alternatively based on an optimization procedure that incorporates terms for cluster quality and novelty relative to existing solution. Once new clusters are detected,outlier factors can be estimated from a new definition for outliers( cluster based outlier), which provides importance to the local data behavior. Experiment shows that the proposed algorithm can detect outliers which exist in different clusterings effectively even in high dimensional data sets.
A novel approach for outlier detection with iterative clustering（ ICOD） in diverse subspaces is proposed. The proposed methodology comprises two phases,iterative clustering and outlier factor computation. During the clustering phase, multiple clusterings are detected alternatively based on an optimization procedure that incorporates terms for cluster quality and novelty relative to existing solution. Once new clusters are detected,outlier factors can be estimated from a new definition for outliers（ cluster based outlier）, which provides importance to the local data behavior. Experiment shows that the proposed algorithm can detect outliers which exist in different clusterings effectively even in high dimensional data sets.
基金
Natural Science Foundation Project of CQ CSTC(Nos.cstc2012jjA 40002,cstc2012jjA 40016)
Fundamental Research Funds for the Central Universities,China(No.0216005207016)
