Multi-objective Optimization Of Reviewer Assignment Problem

With the rapid development of economy,science and technology occupy an increasingly important position in people's lives,countries are also gradually taking technological innovation as one of the core competitiveness.In order to support scientific research and technological development,developed countries,such as the United States,Germany and Japan,have implemented Science Fund System.Assigning appropriate reviewers to proposals and keeping a fair assignment result are the crucial part during the whole fund review process.To ensure the assignment result more reasonable,many scholars have designed different objectives for reviewer assignment problem.However,with the coming of the information age,people have more and more social interaction,and there is no research considering the social relevance between reviewers and proposers.In this thesis,social relevance is considered into the objective of reviewer assignment problem,meanwhile,the balance of matching degree is also considered into the objective of reviewer assignment problem for the fairness among proposals.The objective of matching degree between reviewers and proposals is added to establish the mathematic model of multi-objective optimization in reviewer assignment process,and the model is solved by using the exact algorithm of multi-objective optimization problem.The main research achievements include the following four aspects:(1)The assessment method of matching degree and social relevance between reviewers and proposals.Three factors are considered during computing the matching degree,which are reviewers' expertise,code relevance and coverage rate of key words.Three factors are also considered during computing the social relevance,which are coauthor or teacher-student relationships,region relevance and social association connection between reviewers and proposers.Finally,AHP is used for computing the weight of different factors.(2)Two mathematic models of multi-objective reviewer assignment problem is established when there are many reviewers and there are limited reviewers.When the number of reviewers is large,the mathematic model of multi-objective single-group reviewer assignment optimization is established.When the number of reviewers is limited,the mathematic model of multi-objective multi-group reviewer assignment optimization is established.The objectives of the two multi-objective mathematic model are maximizing the sum of matching degree,minimizing the sum of social relevance and keeping the matching degree of different proposals balance.(3)The multi-objective reviewer assignment problem is solved by using the exact algorithm of multi-objective integer programming.Based on the theory of multiobjective optimization,a more effective algorithm of multi-objective integer programming is proposed for solving the multi-objective reviewer assignment problem.The exact algorithm is used to solve the three-objective knapsack problem to prove its universality and effectiveness,and the complexity of this algorithm is lower and the speed of this algorithm is faster than other exact algorithms when solving the same problem.(4)Numerical experiments and analysis.A real case is used to illustrate the importance of the social relevance in reviewer assignment process.Random data is generated to design experiment of multi-objective reviewer assignment and the models based on the generated dataset are solved by the exact algorithm of multi-objective integer programming.The results show that the exact algorithm can get all the Paretooptimal point of multi-objective reviewer assignment problem within an acceptable time.