Research And Design Of Hybrid Recommendation Based On Explicit And Implicit Feedback

With economy development and science technological progress,Internet data traffic has grown exponentially.The convenience of online shopping has prompted the birth of various e-commerce platforms.Traditional e-commerce systems only gather a large number of commodities,and users can not find the products of interest in timely,effective and accurate manners.The recommendation systems,which predict user preferences by capturing their behavior,can significantly improve the recommendation effect and the user's shopping experience.Most of the existing recommendation systems only use explicit feedbacks clearly expresseing user preferences and degrees,or implicit feedbacks such as clicks,browsing time,which fully records user behavior and has greater authenticity and data volume.The complementary relationship between explicit feedbacks and implicit feedbacks needs to be further explored.And this thesis focouse on improving the recommendation effect by exploiting the heterogeneous feedback data at the same time.The main work of this thesis is as follows:Firstly,based on analyzing the basic architecture,algorithm and user feedback of recommendation systems,a hybrid recommendation algorithm is proposed,which uses both explicit and implicit feedback.Considering the heterogeneity of the feedback data,independent algorithms is uesed to process two kinds of behavior data separately,and then the obtained recommendation result lists are cross-mixed to realize the complementary effect of two kinds of feedback recommendations.Secondly,in order to make full use of user feedback,when processing implicit negative feedback,the maximum similarity is added between the user's unclicked items and clicked items to affect the confidence calculation.The proposed method has higher reliability when compared with the exsisting methods that set the confidence of all unclicked items to the same value.Finally,the proposed model is trained and tested based on the public Movielens ml-100 k data set,with addition of different types of implicit feedback data.Experimental results show that the proposed hybrid algorithm can effectively improve the recommendation effect compared with the classic algorithms,with respedt to the index F1 based on accuracy and recall rate.