Research On Recommendation Algorithm And Recommendation System For Production And Research

As the government vigorously promotes the cooperation of the production and research,it has become an urgent problem about how to establish the links of enterprises and schools,research institutes.How to transform the research results of researchers has become an urgent problem to be solved.It is often time-consuming and inefficient to only rely on the enterprises to search the Internet to meet their needs of the researchers.This thesis focuses on how to research and analyze the implementation of the text representation model and similarity calculation,which are based on the papers published by experts.Prior to the implementation of expert recommendation,this thesis first studies how to select the text representation model and how to calculate the similarity.Clustering and recommendation have a common characteristic,they all need to calculate the similarity,but the clustering results have a variety of evaluation methods,and the recommended evaluation requires historical data's support.Based on the results of clustering,this thesis chooses an optimal text representation method and applies the text representation method to the cold start of recommendation.The main work of this thesis is as follows:(1)Experts' abstract data sets and enterprises' requirements data sets are obtained,and expert's abstract data sets are derived from the virtual reality field for nearly five years,and these data are crawled from the CNKI.Enterprises' requirements data sets are from the Kunshan science and Technology Bureau,commissioned by the three helix research university big data service center,from which this paper selects 500 virtual reality field of enterprise demands.(2)The recommendation of texts involves the text representation and text similarity calculation,this thesis conducts the research on text representation model for comparison,and analyzes the traditional text representation model,and puts forward the representation method of texts which is represented by content based vector and semantic based vector,Then the thesis combines the similarity results of different vector models by linear weighting,and the clustering result using this method is better than the clustering result using traditional method.(3)In this thesis,the above problems are studied by experiments.The thesis compares the results of different models according to the clustering.The experimental results show that the clustering effect of using text representation of TF-IDF&LDA is the best,but the computing time is longer.The clustering effect of using Word2vec-lda is slightly worse,but the computing time is very short.(4)Through the above research work,this paper obtains the best text representation model,and applies the text representation model to the expert recommendation,and calculates the F value of the recommendation result,and evaluates the result of recommendation.(5)The expert recommendation system is designed and implemented.The system can output the relevant experts on the input interface of the recommendation system.Finally,the expert recommendation system is designed and implemented.The experiment shows that the system is feasible and effective.