Design And Implementation Of Internet Novels Recommendation System Base On Hybrid Recommendation Algorithm

With the rapid development of Internet technology,getting information on the Internet has become easier and easier,which change our lifestyles dramatically.At the same time,Internet electronic novels have been rapid developed too.However,there are many kinds of Internet novels and large amounts of books,which makes it difficult for users to find their favorite novels.How to filter the users' current favorite novels list out has become more and more urgent for a novel site.Recommendation systems as a tool to solve the information overload problem have been applied in many fields,but they are rarely applied to the field of Internet novels.In addition,current recommendation systems for Internet novels still exploit the method traditional book recommendation system used,which did not consider the characteristics of Internet novels.Besides,the traditional algorithm need more and more hardware resources due to the increasing quantity of user and product,which means it is difficult to be applied to the production environment.In conclusion,the main works of this article are as follows:1)To solve the precision problem of user rating and users' interest drift,this paper first proposed a method to construct fine-grained rating rules from the chapters,and then introduced time-based user behaviors decay mechanism which improved the ability of the system to capture the change of user interest.A kind of item popularity punishment method is adopted in this paper,which can provide more innovative information and improve the system's diversity.2)This paper propese a hybrid algorithm combining content-based algorithm and tag-ranking-based collaborative filtering algorithm,which avoid the defects of a single kind of algorithm.Tag-ranking-based collaborative filtering raises the ranking of novels which meet the short-term preferences of users by extending the user's preference tag and fusing the extended vector and initial recommendation list.The content-base recommendation algorithm enhances the content relevance of the recommendation result,and solves the poblems of item cold starting and matrix sparsity.3)This paper designed and implemented the Internet novel recommendation system based on the above algorithms,which use the Hadoop file system to store massive user behavior data and novel text data and use the Spark distributed computing framework to complete the daily recommendation task.The proposed approach is validated by a large amount of experiments.The recommended system can be built on the massive data efficiently and accurately.