The Design And Implement Of Implicit Feedback-based Book Recommendation System

In the background of Internet plus,Personalized recommendation systems meet the different demands of consumers by providing them with personalized recommended services.It makes the consumption pattern more interactive than ever and has become an important means of implementing consumption pattern of Internet plus type.However,with the increasing scale of recommended services,some problems narrow the bottlenecks in the traditional recommendation algorithms,such as “Insufficient Rating Data” and “Sparse User-item Rating Matrix”.Therefore,the study of analyzing the user behavior logs and mining the potential interests of the users has attracted more and more attention of academia and industry.And some research scholars try to make up for the deficiency of explicit data by digging out the interest preference of the users.In this paper,in view of the above issues and “One Class Collaborative Filtering”,an improved recommendation algorithm has proposed,aiming to improve the user experience.An algorithm named Hierarchical Implicit feedback Bayesian Personal Ranking has been proposed in a creative way.According to the characters of implicit feedback data,this algorithm uses them as a source to build and update user interest profile.And then a list of recommendation will be got by forecasting the relative preferences on items for users.The main work we have done is summarized as follows:Firstly,the differences between explicit and implicit feedback data and their characters are introduced in this paper.And the paper elaborated the reason of choosing implicit feedback data and the problems caused by the characters of implicit data such as single and difficult to quantize and so on.Then this article introduces some kinds of classical rating-based and ranking-based recommendation algorithms.There are some problems existing in contemporary ranking-based algorithms in which only the differences between “visited item” and “un-visited item” are considered and the degree of preference is lack of understanding and distinguishing.In this paper,the implicit feedback is divided according to the different degree of user interest after a deep exploration of them.In other words,this algorithm uses implicit feedback based on hierarchy model to express user preferences.Based on this,this paper builds a three-layer model for implicit feedback data on Bayesian personalization ranking algorithm,and presents a modified algorithm which can give users recommendations.Secondly,combined with experimental research,the influence of some key parameters on is discussed.Then we compared the new algorithm with the classical method in several evaluation criteria such as Area Under Curve abbreviated AUC,Mean Average Precision,Normalized Discounted Cumulative Gain,Mean Percentage Ranking.The experiment concluded that the new algorithm proposed in this paper could access user's interests exactly.Finally,a book recommendation system based on hierarchical implicit feedback has been designed by analyzing functional requirement and implemented by using the technologies like Java Web,Python and so on.