The Design And Realization Of Examination Paper Auto-generation System Based On Genetic Algorithm

In the information era, the application of new technology in school education promotesthe updating of both teaching means and teaching methods. The traditional way of papergeneration annually is not objective and scientific enough, and questions in the paper and testpaper information are difficult to keep. Therefore, it’s urgently needed to have the traditionaltest methods improved, which can realize the dynamic updates of paper information storageand the test paper database and can free teachers from test paper making.Test paper generation is in essential the multi-objective optimization in a certainconstraint conditions. Under the guidance of Classical Test Theory, the intelligent mathematicmodel of the paper generation is founded on the following basis: analysis of properties andparameters of the test paper, unified quantification of the properties of different attributesaccording to the test score of constraint relation, and realization of the multiple objectiveoptimization solution into single objective one. Based on the analysis of the related literatureat home and abroad, this paper compares and analyses the advantages and disadvantages ofdifferent methods of test paper generation: selection at random, backtracking heuristics andSimple Genetic Algorithm, and chooses the Genetic Algorithm (with ability of globaloptimization) as the method. To avoid repeated questions and premature convergence of thealgorithm in the process of optimizing, this paper uses Improved Genetic Algorithm．It adoptsimproved and segmental real encoding，putting the same type of test items in the samesegment．At the same time, it employs population-sorted based best reserved strategy,avoiding the loss of better individuals. Then repeated questions in the paper are ruled out ofthe paper after several rounds of crossover and mutation operation. Finally, according to theusers’ requirements, the author designs an intelligent paper generation system. The resultsshow that the system can generate reasonable structure of the examination paper and meetusers’ requirements.