Research On Recommendation Algorithm Based On User Dynamic Interest Model

Traditional recommendation algorithms mainly focus on how to connect user's interests with items.Most of them do not take into account the dynamics of user's interests.But in real life,user's interests may change over time.In addition,user's implicit feedback behavior also hides user's interest.Facing with the problem of explicit feedback data sparseness,the recommendation effect of combining it with implicit feedback data that can reflect users' interest is better than that of using only user's explicit feedback data.In this dissertation,we study the dynamics of user interest and implicit feedback behavior information in recommendation system.The main contributions of this dissertation are as follows:Firstly,with the truth of the static user interest model is often used in the current recommendation system,once the user interest change,the recommendation effect is poor,so a modeling method of user dynamic interest model is proposed based on the full study of user interest changes with time.In order to describe the impact of time migration on user's interests,the interest drift model is proposed to update the user dynamic interest model.Secondly,according to the phenomenon of explicit feedback data sparseness in existing recommendation scenarios,a time-aware implicit scoring model is given,which uses implicit feedback information such as time,behavior time,behavior duration and behavior integrity of user behavior to construct the user interest model.Thirdly,based on the user dynamic interest model,a user collaborative filtering recommendation algorithm based on user dynamic interest model and an item collaborative filtering recommendation algorithm based on user dynamic interest model are constructed.These algorithms are based on the user's dynamic interest model that are able to provide users with more accurate and personalized services.Fourthly,due to each recommendation algorithm has its own advantages and disadvantages,based on the advantages of various recommendation algorithms when generating recommendation lists,a hybrid recommendation algorithm is proposed based on user dynamic interest model by using cascade method and switch method of hybrid recommendation algorithm,which can achieve accurate recommendation and enhance user experience and user viscosity to the system.Last but not least,the three proposed recommendation algorithms based on user's dynamic interest model and other original algorithms are compared and analyzed by using real datasets in IPTV domain.The experimental results show that the proposed recommendation algorithms improve the accuracy of recommendation and has stronger practicability.