A Novel Multidimensional Scaling Technique for Mapping Word-Of-Mouth Discussions
The techniques which utilize Multidimensional Scaling (MDS) as a fundamental statistical tool have been well developed since the late 1970’s. In this paper we show how anMDS scheme can be enhanced by incorporating into it a Stochastic Point Location (SPL) strategy (one which optimizes the former’s gradient descent learning phase) and a new Stress function. The enhanced method, referred to as MDS SPL, has been used in conjunction with a combination of the TF-IDF and Cosine Similarities on a very noisy Word-Of-Mouth (WoM) discussion set consisting of postings concerning mobile phones, yielding extremely satisfying results.
KeywordsMobile Phone Multidimensional Scaling Stress Function Cosine Similarity Multidimensional Space
Unable to display preview. Download preview PDF.
- 3.Oommen, B.J.: Stochastic searching on the line and its applications to parameter learning in nonlinear optimization. IEEE Transactions on Systems, Man and Cybernetics SMC-27B, 733–739 (1997)Google Scholar
- 6.Oommen, B.J., Granmo, O.C., Liang, Z.: Mapping Literature and Word-Of-Mouth Discussions Using Multidimensional Scaling, Unabridged version of this paper (submitted for publication) (2008)Google Scholar