Font Size: a A A

Research And Application On Collaborative Recommendation Method Of Project Review Experts

Posted on:2017-08-06Degree:MasterType:Thesis
Country:ChinaCandidate:W X YangFull Text:PDF
GTID:2348330482476794Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
In recent year,with the increasing number of science and technology project application,the project review task became increasingly onerous.Basically,project evaluation experts were artificially assigned by the management by subjective consciousness or extracted randomly from the system according to the research field.Therefore,how to efficiently select the best matching project evaluation experts has become an important issue in the field of science and technology project management.To this end,this paper mainly studies the key technologies of science and technology project balanced grouping method based on topic,the combination of content and collaborative filtering recommendation method of expert,expert scoring model and recommendation algorithm optimization and so on.The main work is as follows:1.In order to facilitate the arrangement of the project group review,we build science and technology projects' knowledge model and propose a project balanced grouping method based on LDA topic model.First of all,we mine the underlying topic distribution in the set of projects using LDA topic model,build the project-topic complete bipartite graph according to the related topic and realize the initial grouping of project.In the case of project number of group differ greatly,put forward a balanced grouping strategy of project based on balanced degree and discrimination of topic.Traversal project in large group according to the rules and parameters setting and fine-tuning the individual projects.Then we achieve the compromised grouping result of similar topic and balanced quantity.2.Related to the experts' evaluation feedback,we propose an experts' recommendation method for project evaluation based on combination of content and collaborative filtering.First of all,we propose a similarity computing method between project topic vectors and expert vectors using improved cosine formula,and complete the recommendation algorithm based on content.We build a matrix of experts' interest based on evaluation feedback and achieve the collaborative filtering recommendation method combining the project review information in history.Using linear combination strategy,we propose the expert' recommendation method fusing content and collaborative filtering.3.We propose an expert scoring model and optimized expert' recommendationalgorithm.We define weight set based on experts' positions,analyze the attitude of experts based on scores in historical examination record and results of project application,and build expert scoring model.We optimized the expert'recommendation algorithm considering the factors such as the corresponding of expert and project,ability of academic research,and record of credit,by the way of weighted factor to score compositely.Our method makes the effect satisfied real requirement.Finally,we develop a project examination expert' recommendation system based on the results mentioned above,and have a try on Zhejiang scientific project application system.The system assists the department of project arrangement to arrange expert examination scientifically and effectively,and verify the effectiveness of our research results.
Keywords/Search Tags:collaborative recommendation, feedback, balanced group, potential topics, LDA, scoring model
PDF Full Text Request
Related items