A Study Of Generalized Assignment Methods For Team Recommendation

Team recommendation is a member selection problem that requires both a clear division of labor and cooperation and has comprehensive application value in many fields,such as resource management.How to construct team recommendation methods with interpretability and operability is still a wide research concern in academic and application fields today.This thesis addresses the team recommendation problem,and on the basis of systematic analysis of the essential features of team recommendation and the shortcomings of existing methods,the following work is mainly done:(1)a formal description of the team recommendation problem is given,based on the work ability values of alternative members,the integrated work ability value of the team as the core concern indicator,and the importance weights of tasks and the energy allocation weights of members as the data fusion strategy,a generalized assignment model(TRGAM)for team recommendation is established,and the basic properties and standardization measures of TRGAM are discussed;(2)as an application of TRGAM,a system reliabilitybased team recommendation model(BRTRM)was developed based on the work capability values of alternative members,the reliability of the team in completing the whole work as the core concern indicator,and the capability enhancement operator as the complementarity measure among members;(3)the solution strategies E(?)H-TRGAMs and E(?)H-BRBTRM with enumeration method and Hungarian algorithm as integrated algorithms are given for the standard forms of TRGAM and BRTRM(TRGAMs and BRBTRM),and the effectiveness of the algorithms is tested with specific case studies;(4)for the large-scale team recommendation problem,the solution algorithm G(?)H-TRGAMs with genetic algorithm and Hungarian algorithm as local optimization is designed for TRGAMs,and the solution performance of G(?)H-TRGAMs is further analyzed with specific cases.Theoretical analysis and experimental results show that TRGAM and BRTRM have good structural characteristics and interpretability.Their solution algorithms E(?)HTRGAMs,G(?)H-TRGAMs and E(?)H-BRBTRM all have good solution performance.These discussions enrich the existing related theories and methods to a certain extent and have specific theoretical and application values.