Collaborative Filtering Recommendation Model With Explicit And Implicit Feedback

With the explosive growth of information scale,Recommendation System(RS)can help people easily find the information they are most interested in.Collaborative Filtering(CF),as one of the most effective methods in RS,has been widely studied and applied in academic research and industry.In the RS,the user's feedback mechanism can be divided into two types according to whether it can clearly indicate user preferences: explicit feedback(scoring,evaluation,etc.)and implicit feedback(browsing history,purchase history,etc.).In general,most existing CF recommendation models use one of explicit feedback and implicit feedback to analyze users' preferences and items' characteristics,while not using the different characteristics of both to make more effective recommendations.In addition,most models are designed for rating predictions or item sequencing tasks,and few models can accommodate both tasks simultaneously.Based on the above problems,this thesis proposes a CF recommendation model with explicit and implicit feedback,which can effectively combine explicit and implicit feedback in a unified model.Specifically,the model first uses a deep neural network collaborative filtering model that introduces attention mechanism for implicit feedback.This model uses the attention network to distinguish the importance of different items,replacing the item similarity in the traditional CF recommendation method.The metrics are used to model the high-order nonlinear relationships between users and items through deep neural networks to complete project sequencing tasks.Then,based on the explicit feedback,the probability matrix factorization model is used to complete the rating prediction task of the item,and the candidate items set generated based on the implicit feedback is reordered according to the predicted rating,thereby generating a user recommendation list.The main work of this paper can be summarized as following :Designed the Collaborative Filtering Recommendation Model(EI-CF)with explicit and implicitfeedback to incorporate explicit and implicit feedback.The EI-CF model is aimed at the singleness problem of user feedback mechanism in RS research work.From the perspective of combining explicit and implicit feedback,the CF recommendation model which can complete the rating prediction and project sequencing tasks at the same time is designed.The model first selects the item set that the user is most interested in according to the implicit feedback to complete the project sorting task,then predicts the user's score on the candidate item set according to the explicit feedback,and finally reorders the candidate item set to generate the user recommendation list.Finally,the paper conducts experiments on two public datasets and compares them with the recommended methods representing different categories of RS to verify the validity of the model.The experimental results show that the model in this paper shows better recommendation performance on these two data sets than other methods,which provides a valuable reference for the future development of RS based on the combination of explicit feedback and implicit feedback.