The Research On Algorithm Of Adaptable Paper Generation Based On Hybrid Genetic Algorithm

Now is the era of computer technology and the Internet.The rapid development of science and technology has also greatly changed our way of life.With a variety of academic level tests,recruitment exams,professional skill level tests,etc.gradually transferred from offline to online,a number of online question banks and on-line examination systems have also come one after another,as a core element,the question of test paper generation naturally attracts corresponding attention.However,the existing results often do not meet all the requirements of test paper generation,they simply take one or two of them.But a real good adaptability method of test paper generation should not only generate test paper efficiently,but also necessary to ensure the quality of test paper generation and to meet the user's various requirements,instead of considering only certain aspects,so the method can be more adaptable.Ultimately it can not only be suitable for a variety of different types of examinations,but also have more practical significance in the customization of other personalized teaching resources.Aiming at the existing problem of test paper generation,this paper presents a algorithm based on a hybrid genetic algorithm(SNGA).First start from the origin of the research object,proposed the definition of test paper model,and the problem which involves a variety of constraints has been carefully divided and organized.And convert it into an objective function that contains multiple conditions such as overall distribution,knowledge distribution,and chapter distribution,basically covers all the constraints required by the problem.Then the paper made a corresponding improvement based on the traditional genetic algorithm.First,we designed the method of diversity measurement for the population.Above the most widely used measure of population entropy,added the assessment of the overall genetic similarity of the test paper.In this way,a more accurate measure of the population diversity as a whole is obtained from the aggregation of different fitness and the similarity of the population gene position,it also lays the foundation for adaptive adjustment of subsequent crossover and mutation probability.In order to improve the deficiency of the original genetic algorithm that may fall into the local optimal,multiple niche congestion algorithm is integrated into genetic algorithm,the difference from the general niche genetic algorithm is that the multiple niche congestion genetic algorithm promotes a heuristic conditional pairing mechanism and a random individual substitution mechanism.In this way,the loss of good genes can be reduced at the time ofmating,when replacing old and new,we are not limited to father and son alone.Then based on the measurement of population diversity,an adaptive adjustment of the crossover and mutation probability was proposed.First,adjust the overall population's crossover and mutation probability based on the overall diversity of the population,then adjust the individual probability based on the overall probability.In the early and late stages of evolution,the probability is fine-tuned with the diversity of the population,and further delays the single problem of population diversity in the later stages of evolution.Finally,based on the real data of the team's existing online database,a large number of experiments were conducted on a variety of related algorithms,and through the commonly used detection indicators to determine the experimental results,in consideration of comprehensive indicators,the algorithm of test paper generation proposed in this paper has a slight improvement in the efficiency and the quality compared with the other algorithms in the experiment,basically meet the actual needs.