An improved sIB algorithm for document clustering using combination weighting measures

This paper presents an improved sIB algorithm (CW-sIB) for high dimension document clustering using combination weighting. Traditionally, feature weighting researches on clustering devote themselves to search one single effective weighting scheme. However, how to choose a proper weighting scheme is a generally acknowledged devilish problem. To address this issue, we propose the linear combination weighting method derived from the idea of combination evaluation for multiple attribute decision making problem. The application of combination weighting can overcome the limitations of using single weighting scheme. It will help to reflect the essential characteristics of the document data better. The experiments on real document data have shown that the proposed CW-sIB algorithm is superior to the sIB algorithm. Meanwhile, we report results as to which combination of weighting scheme elements show merit in the decomposition of datasets.