Study On New Modes Of Interactive Group-buying Websites And Group Recommendation Algorithm Development

Recently, there has been rapid development for network group buying in China.However, too rapid development also brings about such accompanying problems asinformation overload, homogenization, and so on. Many websites attempt to addressthese drawbacks but fail eventually. As an effective way to reduce informationoverload, the individual recommendation technique has received a lot of interests.Nonetheless, little attention has been paid to the study of group recommendation. Inview of the benefit that symbolic data analysis can reduce the dimension of mass dataand capture the sample property on the whole, this technique has been widely appliedin recommendation, but it is still in the elementary stage. Motivated by the desire totackle the aforementioned problems, this thesis proposes a new mode of interactivegroup-buying website and then develops a novel group recommendation technique. Inparticular, the main work of this paper is as follows.Firstly, the explicit form and operation mechanism for interactive group-buyingwebsites are proposed. The bidirectional function of network group-buying websitesis investigated and the sense for the term “interactive” is extended in accordance withpractical requirements. Then, some analysis is implemented for the correspondingprofit, cost-effectiveness, future development, and so on. We aim to establish aconsolidated mechanism for group-buying websites.Secondly, the approach to expressing group models is given, by utilizing theconcept of interval and distributed symbolic data. On this basis, the thesis thenproposes a symbolic data-based group recommendation algorithm applicable to thenew mode of group-buying website. By collecting real data from www.dianping.com,some experiments are conducted to verify the effectiveness and precision for theproposed algorithm, which indicates that it can achieve superior performance withrespect to existing methods.