Design And Implementation Of Real-Time Recommender System For Academic And Expert System Based On Flink

With the rapid development of science and technology,academic achievements such as papers and patents are increasing day by day,but at the same time,it has also brought about the problem of information overload.To alleviate this problem,academic and expert systems have emerged,providing rich academic resources and search functions.However,existing academic and expert systems generally suffer from the problem of strong retrieval and weak recommendation.Even if there are recommendation services,their strategies often focus on offline recommendations such as domain topic recommendations,ignoring users’ immediate interests.In addition,the massive academic resources of academic and expert systems also place extremely high demands on the performance of recommendation systems.Therefore,providing high-performance real-time recommendation services that can capture immediate interests for academic and expert systems is of great value and significance.To address these issues,this paper designs and implements a real-time recommendation system for academic and expert systems based on Flink,with the following key work:(1)Design and implement a three-stage Flink-based real-time stream processing recommendation architecture.This architecture effectively integrates the advantages of the Lambda architecture and the Kappa architecture,and is compared with a Lambda architecture scheme based on Storm in experimental results,which verifies that the proposed architecture has significantly improved throughput and latency.(2)Propose a multi-channel recall real-time recommendation strategy for academic and expert systems.This strategy combines offline recommendation strategies such as collaborative filtering and content-based recommendation with real-time recommendation strategies such as popularity-based recommendation and real-time collaborative filtering recommendation,and customizes the fusion strategy according to the characteristics of the academic and expert system’s multi-scene recommendation.Finally,using real user data collected by the platform,ablation experiments and branch comparison experiments verify that the proposed strategy improves the hit rate,recall rate,and NDCG metrics.(3)Build an academic and expert system platform with real-time recommendation.The platform system uses the SpringBoot framework for backend development,Vue.js and Nuxt.js for frontend development,MySQL as a relational database,and is deployed on an Alibaba Cloud ECS server.The platform system implements modules such as homepage recommendation,project and seminar functions,search function,workspace function,and administrator function,with complete functions.After functional and performance testing,the platform system has been launched and is in operation.