Research And Discussion On Collaboration Filtering Arithmetic In Recommendation System Of Active E-Marketing Mode

With the rapid development of information and network technology, we have entered the Internet era. Internet can provide not only a convenient way to collect all kinds of latest information, but also a business platform via network access to the enterprises. It can help the enrerprises to exploit marketing. The enterprises carry out a series of commercial activities via Internet, e-communication and digital exchange which enhance the communication and relationship between the enterprises and its customers. At the same time, the enterprises establish the recommendation system in the current enterprise business network to achieve the active e-marketing mode. In this mode, the enterprises can influence the choices of its customers as well as collect all kinds of information of its customers. As a result, the enterprises can realize the strategic change from product-centered to customer-centered, and on the other hand, the customers can find out their statisfactory products easily.Although the recommendation system has been widely used in e-business and e-library, it is still faced with series of new technical challenges with the enlargement of recommendation system. The paper will research on the collaboration filtering arithmetic which is widely used in the recommendation systm as the problems existing in this system. And the problems and challenges of collaboration are analysed. Then in order to solve the problems of sparity and cold-start in current collaboration filtering arithmetics, improved collaboration recommendation arithmetic is presented in this paper. And a new feature similarity measurement is used to raise the quality of recommendation. And what's more, a multi-functional personalized recommendation system is constructed to meet customer's needs.In this essay, the main content and structure are arranged as follow:Firstly, the developing actuality of recommendation system and the relevant technology are introduced. And the recommendation system's new characteristic under active e-marketing mode is analysed. At the same time the purpose and the significance of this essay are expounded.Secondly, the respective advantages and disadvantages of classical recommend- ation arithmetics which are now widely used in the recommendation system are discussed; the collaboration filtration recommendation technology is analyzed specifically in this paper as well. Then, improved effective collaboration filtration arithmetic is put forward to solve the outstanding problems such as "sparsity"and "cold-start".Thirdly, the paper tests and appraises the improved recommendation arithmetic presented in this paper on using the data-sets of the collaboration filtering authority field. In this way, the improved arithmetic is proved to be rational, effective and creative by the experiment.At last, a multi-functional personalized recommendation system prototype is constructed according to demands of enterprise, combinied with the improved arithmetic. Therefore, it fulfills the practical application of this research according to the real demands.