DCSO: Dynamic Combination of Detector Scores for Outlier Ensembles

Selecting and combining the outlier scores of different base detectors used within outlier ensembles can be quite challenging in the absence of ground truth. In this paper, an unsupervised outlier detector combination framework called DCSO is proposed, demonstrated and assessed for th e dynamic selection of most competent base detectors, with an emphasis on data locality. Th e proposed DCSO framework first defines the local region of a tes t instance by its k nearest neighbors and then identifies the top-performing base detectors within t he local region. Experimental results on ten benchmark datasets demonstrate that DCSO provides consistent performance i mprovement over existing stati c combination approaches in mining outlying objects. To facilitate interpretability and reliability of the proposed method, DCSO i s analyzed using both theoretica l frameworks and visualization techniques, and presented alongside empirical parameter setting instructions that can be used to improve the overall performance.