| Web service applications,such as E-commerce and social networks,have become increasingly mature in the boom of network technologies and information services.The personalized recommendation algorithms play a more and more important role in supporting the web services on the social platforms and business platforms.The collaborative filtering recommendation algorithm has been successfully and widely used in recommended systems,especially in E-commerce,due to its excellent data modelling techniques and intelligent service specifications.However,the increase of the users and items data in recommended system will definitely lead to data sparsity.And many researchers proposed different categories of the collaborative filtering recommendation algorithms based on clustering to ameliorate the situation of the data sparsity.As the most classic clustering algorithms,the K-means algorithms adopt the central point randomly,thus the K-means algorithms may not achieve the optimal results.At the same time,when calculating the similarities by using recommendation algorithm,people may not get good recommendations because the calculation is based on ratings of user-item,without considering the features of the user and the item itself.In order to address the aforementioned issue,this work proposes an enhanced collaborative filtering recommendation algorithm.Specifically,the paper makes the following contributions:At first,the paper detailedly introduces the research background and some theories about the traditional recommended system.We also cover the content of the traditional recommendation systems based on which we proposal our study.Then,the paper analyzes the problems of the conventional K-mean and Collaborative filtering recommendation algorithms,and proposes the enhanced algorithm.Our algorithm enhances the traditional K-mean clustering recommendation approaches in two aspects: 1)choosing deterministic clustering central point and clustering process;2)considering user features and item properties.In the end,the author compares the performance between the traditional and the enhanced algorithms.It is demonstrated by the experiments that the enhanced algorithms could efficiently improve the quality of the recommendation systems. |
