A Case Study of Core Vector Machines in Corporate Data Mining

Abstract

The core vector machine (CVM) has been introduced as an extremely fast classifier which is demonstrably superior to standard support vector machines (SVMs) on very large datasets. However, only limited information regarding the suitability of CVM for supporting corporate planning is available so far. In this paper, we strive to overcome this deficit. In particular, we consider customer-centric data mining which commonly involves classification in medium-sized settings. CVMs are compared to SVMs within the scope of an empirical benchmarking study to clarify whether previous findings regarding the competitiveness of CVMs generalize to business applications. To that end, representative real-world datasets are employed. In addition, the study aims at scrutinizing the behavior of CVM during model selection. Following a standard grid-search based approach we find some evidence for CVM being more sensitive towards parameter settings than SVMs.