Research On Two-way Recommendation Of Graduate Employment Based On Collaborative Filtering And Thematic Model

With the continuous increase in the number of college graduates,the employment of college graduates has received more and more attention.College graduates often face problems such as unclear goals,inaccurate positioning,and confused psychology.Many university graduates take the resume distribution strategy of "spreading the net and focusing on fishing",and cast their resumes.This not only brought great trouble to the graduates themselves,but also brought a lot of inconvenience to the company.Companies face a lot of recruitment options,often need to screen the graduates,leading to the company's high labor costs.In view of the above,this paper proposes a mixed recommendation model for graduates and businesses to achieve two-way recommendation,using the A-recruitment website as a platform,combining graduates'resume data,resume delivery records,the name of the company,recruitment positions,and cities involved.The recommendation of target companies to graduates can reduce the confusion of graduates in the process of finding a job.For enterprises that recommend target graduates,enterprises can directly send interview invitations to target graduates to directly contact and reduce the cost of the company in the recruitment process.Therefore,it is very beneficial to both companies and graduates.This article mainly adopts methods such as improving collaborative filtering algorithm and LDA theme model to realize two-way recommendation for enterprises and graduates.First of all,we preprocess the collected data,including normalization,word segmentation,and stop words.Then build a corresponding matrix,using the user's resume delivery data and corporate interview invitation data.Then we adopt the K-means clustering method to solve the problem of data sparsity in the matrix,and then adopt an improved collaborative filtering algorithm to process and implement recommendations.After the data is preprocessed,we use the LDA topic model to obtain the subject word information and the TF-IDF method to determine the weight of the topic word.Then we establish the vector space model of the graduates and the company combine the recruiting positions of the recruiting companies,involving the city,the recruitment brochure information,etc.,then calculate the similarity of the vector space model,and implement the recommendation.Finally,a hybrid recommendation model is established by combining two methods.In the test experiment,we first use the user's satisfaction data for the recommendation to determine the coefficients in the hybrid recommendation model.Then,the relevant data is obtained by setting up a control experiment,including the changes in the number of resumes submitted by graduates,the changes in the number of interviews sent by companies,and the time data for companies to fill vacancies.The comparison of the three algorithm recommendation effects shows that the effect of the mixed recommendation is the best conclusion.Finally,we use the activity of the website(average user online time),the number of new users,user satisfaction with the website,the average number of resumes submitted by graduates,and the average number of interview invitations sent by companies to verify the effectiveness of recommendations.Experiments have proved that the recommendation is effective and very beneficial to the development of the website.