Design And Implementation Of Campus Association Activity Recommendation System

With the rapid development of Internet technology,artificial intelligence technology and big data,it ushered in the 2.0 era of educational informatization characterized by the application and innovation of new technology.How to construct the intelligent campus of wisdom education,teaching and management has become one of the important topics of university education and management research.Specifically,student associations,served as an important extension of “the first lesson” and a core carrier of “the second lesson”,play a significant role in achieving the popularity of higher education.However,statistics shows that the student clubs' marketing promotion,management and the participation approaches of group activities heavily depend on the inefficient manual interventions in many domestic universities.This leads to unsatisfactory marketing results,inadequate customized group events,and time-consuming and labor-intensive office tasks.This article aims to explore the necessity of student associations in university life given the above background,and to address two issues through researching on the personalized recommendation algorithms: the data sparseness problem in collaborative filtering algorithm and the semantic analysis flaws found in the traditional content-based recommendation algorithm.Additionally,this thesis designs and implements a recommendation system for university student association activities,through improving both content-based recommendation and collaborative filtering algorithms under group events scenarios.The main structure of this thesis is listed below:1.Based on Jaccard coefficient to measure similarity,sparse matrices of users and item-scoring in collaborative filtering algorithm normally lead to the quality issues during practical system recommendation process.Hence,this thesis improves this similarity measurement by incorporating a heat penalty term and a weighted common-scoring factor,leading to a more accurate user similarity matrix.First,thismatrix can be obtained by reducing the effect of popular items on the recommendation outcomes via the heat penalty term.Besides,the user similarity measurement process will analyze the impact of user scoring differences and the relationships between common-scoring factors and all-scoring factors.2.This thesis aims at the shortcomings of the traditional content-based recommendation algorithm in semantic analysis and proposes to weight the text similarity calculated word2 vec word vector with the original TF-IDF-based community similarity,and obtains the improved algorithm of community recommendation based on TF-IDF and word2 vec to improve the accuracy of system recommendation.At the same time,this thesis further compares the revised collaborative filtering algorithm and content-based recommendation algorithm mentioned above with traditional algorithms,using real data from Meetup(a largest service that organize online groups and host in-person events outside of China).These comparisons demonstrate that the improved algorithms introduced in this thesis alleviate the cold start problem of new users adding in the system,and better the recommendation results when having sparse data.3.Last but not least,this thesis optimizes the operational process of student association events and thoroughly evaluates the demands and designs of recommendation systems,including system architecture,functional structure,interface design and database structure.This enables the system to recommend customized group events to university students.