Less is More: Building Selective Anomaly Ensembles
This work addresses the challenge of improving anomaly detection accuracy in temporal graphs, such as email communications and social media feeds, by introducing a selective ensemble approach, though it is incremental as it builds on existing ensemble techniques.
The paper tackled the problem of building effective anomaly ensembles for event detection in temporal graphs by proposing SELECT, an unsupervised method that selectively combines results from heterogeneous detectors, achieving superior performance compared to individual detectors and other ensemble approaches on real-world datasets.
Ensemble techniques for classification and clustering have long proven effective, yet anomaly ensembles have been barely studied. In this work, we tap into this gap and propose a new ensemble approach for anomaly mining, with application to event detection in temporal graphs. Our method aims to combine results from heterogeneous detectors with varying outputs, and leverage the evidence from multiple sources to yield better performance. However, trusting all the results may deteriorate the overall ensemble accuracy, as some detectors may fall short and provide inaccurate results depending on the nature of the data in hand. This suggests that being selective in which results to combine is vital in building effective ensembles---hence "less is more". In this paper we propose SELECT; an ensemble approach for anomaly mining that employs novel techniques to automatically and systematically select the results to assemble in a fully unsupervised fashion. We apply our method to event detection in temporal graphs, where SELECT successfully utilizes five base detectors and seven consensus methods under a unified ensemble framework. We provide extensive quantitative evaluation of our approach on five real-world datasets (four with ground truth), including Enron email communications, New York Times news corpus, and World Cup 2014 Twitter news feed. Thanks to its selection mechanism, SELECT yields superior performance compared to individual detectors alone, the full ensemble (naively combining all results), and an existing diversity-based ensemble.