Research On Explainable Personalized Recommendation Algorithm Based On Academic Social Network

With the continuous development of academic social network platforms,digital resources such as information consultation and teaching videos have shown an amazing growth rate,how to effectively and accurately find useful information from massive digital resources according to the interactive behavior data between learners and information.Explainable personalized recommendation systems have become a key technology to activate the potential of academic social data elements,and personalized push can also improve the coverage of learners’ retrieval information and provide different services for different types of learners.The classic traditional recommendation algorithm is widely used in the industry because of its simple implementation ideas and strong stability.However,such algorithms usually have problems such as cold start,unexplained reasons for recommendations,and single application scenarios.In order to solve such problems,this paper relies on the academic social networking platform "Scholat.com" to carry out research work,and the specific contributions are as follows:1.Aiming at the cold start node problem in Scholat.com,a personalized explainable learner implicit friend recommendation method(PELIRM)is proposed.The main idea of the PELIRM algorithm is,firstly,to obtain the trust between learners through the potential friend mining method of three-degree influence,and judge the similarity of common preferences between the two according to the research interests of learners.Secondly,the learner’s check-in information is used to determine the spatial distance between learners,and learners in the same city or neighboring areas are screened out as a recommendation list.Finally,with the interpretability of PELIRM,the trust in the recommendation effect of cold start nodes is enhanced.Experimental results show that the recommendation effect of PELIRM algorithm is significantly improved compared with other potential friend mining algorithms on Scholat.com datasets.2.Aiming at the recommendation problems such as the order of Scholat.com recommended content and the correlation of learners’ input keyword information,an explainable personalized recommendation method based on self-attention mechanism(EPRSA)is proposed.The main idea of EPRSA algorithm,through the learner’s historical query sequence and historical click,will extract the topic preference information in the learner’s historical click document,and make it the attention information of the LSTM(Long Short-Term Memory)model through the self-attention mechanism,and predict the learner’s click preference for the input question answer.Add explainable justifications to increase the accuracy of recommended content through EPRSA prediction of learners’ click preferences.Experimental results show that the recommendation effect of EPRSA algorithm is significantly improved compared with the traditional algorithm on Scholat.com datasets.