Study on Personalized Recommendation Technology Based on Indirect Uncertain Data
|Course||Applied Computer Technology|
|Keywords||Personalized recommendation User-Item missing ratings Ratings Complement Two-dimensional normal distribution Fine-grained recommendation|
With the rapid growing of the number of the users and the items in e-commerce site, extreme various User-Item transaction data and large numbers of meta-data describing the users and the items, which are often based on uncertain data, appear. The uncertain data of e-commerce hinder improving the recommendation quality and increasing customer’s satisfaction. Therefore, there are important theoretical and practical significance in the study of uncertain data in e-commerce site.First of all, this paper introduces the e-commerce, personalized recommendation technology, and uncertain data and so on. At the same time in order to meet researching the uncertain data of e-commerce needs, this paper summarizes uncertain data of e-commerce. At the same time, the uncertain data of e-commerce are classified into the direct uncertain data and the indirect uncertain data. And then analyzes the emergence causes of all kinds of the uncertain data.Secondly, a detailed analysis is done on the User-Item missing ratings of the e-commerce site, which are very important in the implementation of the personalized recommendations. Based on the statistical and the analysis of the uncertain data, to solve the issue which considers the ratings factors is unilateral in the existing methods, the missing User-Item rating complement model (UIRCBTRV) based on the two-dimensional random variable which is two-dimensional normal distribution is proposed. On the basis of the model, the two-dimensional User-Item rating complement algorithm (UITRC) is designed. And then analyze the complemented accuracy and recommendation quality of the two-dimensional User-Item rating complement algorithm (UITRC).Thirdly, the existing individual recommendation system, which recommendation results are the user set and the item set based on the model of many users to many items, are mostly coarse-grained and imprecise recommendation. Therefore, the Fine-grained model (FGRM) is established, and the Fine-grained recommendation algorithm (FGRA) is designed, which improves the recommendation quality.Finally, the proposed models and algorithms are inspected with experiments. The experimental results show that the two-dimensional User-Item rating complement algorithm (UITRC) and Fine-grained recommendation algorithm (FGRA) can be in line with the distribution of the User-Item ratings and the expected results of the precise recommendations based on the two models proposed in this paper. By comparison with the existing algorithms, the models and the algorithms proposed in this paper can improve the recommendation quality of recommendation system.