KBSVM: KMeans-based SVM for Business Intelligence
Jiaqi Wang, Chengqi Zhang · Journal of the Association for Information Systems · 2004
The goal of business intelligence (BI) is to make decisions based on accurate and succinct information from massive amounts of data. Support vector machine (SVM) has been applied to build the classification model in the field of BI and data mining. To achieve the original goal of BI and speed up the response of real-time systems, the complexity of SVM models should be reduced when it is applied into the practical business problems. The complexity of SVM models depends on the number of input variables and support vectors. While some researchers have tried to select parts of input variables to build the model, this paper proposes a new method called KMeans-based SVM (KBSVM) to reduce the number of support vectors. The experiments on real-world data show that the KBSVM method can build much more succinct model without any significant degradation of the classification accuracy.