Research On Technology Of Personalized Recommendation In E-business Based On Data Mining

As Internet spreads among the people, electronic business has developed greatly owing to such characteristics as convenience with rapidness, high efficiency with low cost. The scope of the business web sites has become increasingly extensive and its structure has become more and more complicated. Consequently, on the one hand, faced with large amount of commodity information, customers often hardly find out the goods that they need smoothly. On the other hand, enterprises are eager to solve such problem as how to improve attraction of their web sites efficiently to raise the level of customer service and finally to gain more business profits. Recommendation system of electronic business is the efficient method to solve this problem.After the thesis analyzes and examines thoroughly the relevant theories and key techniques about the current recommendation system, it points out some problems among the existing recommendation system. Then the thesis proposes a structure of recommendation system of electronic business and introduces the functions and constructions of each module in the system in greater detail.The thesis also puts forward CPPICF based on consumers'favorites to improve accuracy of the individualized recommendation system which focuses on the problem of customer contiguity in coordination filtering and realizes subjective evaluation of information to solve the problem of data sparseness better and has been tested through experiments. What's more, the author of the thesis makes a design of CRF to correct any deviations in recommendation effects and makes an evaluation about practicability and validity of the algorithm, Enhancement the customer degree of satisfaction, improved the recommendation quality.The shortcomings of the thesis lie in that operation efficiency of CPPICF is not high when the amount of data is large enough and individualization of recommendation effects is not apparent because more super chain recommendations are employed. Therefore, the successive researches will focus on transforming the model of customers skimming web pages into the model with higher representative ness and less data to improve the algorithm. On the other hand, the author of the thesis will conduct researches on how to adopt the suitable bound forms based on customers'interests (e.g. drifting advertisements etc.) to provide recommendations for customers in order to improve individualization.