Multi-objective Optimization Method For Large-scale Evaluation Problems

The research on multi-objective optimization method of evaluation problem has very important theoretical significance and application value.Because the evaluation object has individual difference,it is one of the largest key problems in large-scale evaluation that how to classify the evaluation object reasonably,and then evaluate the objects scientificallly according to the classification results.Firstly,this paper improves the clustering algorithm proposed by Tzortzis and Likas,proposes a new improved K-means algorithm,and then we use the improved K-means algorithm to classify the objects in large-scale evaluation,by introducing the score conversion function and satisfaction function,a multiobjective optimization model is built to study the large-scale evaluation problem.Chapter 1,in this paper,the multi-objective optimization problem,K-means algorithm and evaluation problem are presented.Chapter 2,this paper mainly improves the Minmax K-means algorithm proposed by Tzortzis et al.Because the Minmax K-means algorithm uses Euclidean distance to describe the differences between individuals,it is only applicable to the individual clustering problem of the absolute value of individual evaluation index.However,the individual differences caused by the different professional backgrounds of evaluation objects are not only reflected in the absolute value of each index,it is also reflected in the bias of each index value of the evaluation object.Therefore,the classification of the evaluation object only considers the absolute difference of each index value,which may lead to unreasonable classification.In this chapter,by introducing cosine distance,Cosine similarity is defined to describe the difference between evaluation objects.A new distance function is constructed by Euclidean distance and cosine similarity to describe the similarity between evaluation objects,and the algorithm proposed by Tzortzis is improved.Chapter 3,the improved K-means clustering algorithm is applied to large-scale evaluation problems.Because of the large scale of evaluation objects in large-scale evaluation problems,if we study each evaluation object,the model constructed is generally very complex,which will bring great obstacles to the efficient solution of the model.In this chapter,firstly,the evaluation objects are classified reasonably,and the score conversion function is introduced,the workload of each evaluation object is divided into the basic workload and the workload of participating in performance evaluation.Secondly,the most favorable and disadvantageous optimal allocation model of each evaluation object is constructed to obtain the most favorable and disadvantageous evaluation scheme of each category.Based on this,the average satisfaction function of each category is constructed based on the classification result of evaluation object,the multi-objective optimization model is constructed to maximize the average satisfaction of each category.The numerical experiment results show that the improved K-means algorithm proposed in this paper can be effectively used for the classification of evaluation objects in large-scale evaluation problems,and the multi-objective optimization model constructed for large-scale evaluation problems has better numerical experiment results.