Item Selection Strategy With Quantum Intelligence Algorithm And Item Response Theory

In order to explore the performance of the item selection strategy by the quantum intelligent algorithm , this paper will compare Genetic Algorithm(GA) and the Quantum Genetic Algorithm (QGA), Particle Swarm Optimization (PSO)Algorithm and Quantum Particle Swarm Optimization(QPSO) algorithm, Ant Colony Algorithm(ACA) and Quantum Ant Colony Algorithm (QACA) pairwise.This study is based on the simulated item bank , using Item Response Theory of the three parameters logistic model to establish the objective function of each algorithm, ANOVA method is used to analyse the parameters which affect the result of test assembly,and to find the optimal parameter combination.Finally,choose the best algorithm from the three pair of algorithms.The results enriched the theory of the current item selection theory, and applied the Quantum Intelligent Algorithm in the test assembly successful.Methodology, it's not only a breakthrough in the field of item selection, but also for the application of artificial intelligence in psychological measurement. The main research conclusions are the following:(1) The results of Genetic Algorithm for this item selection experiment showed that although the test information on the cutoff score is large, but after several times item selection experiment the standard deviation is greater, so the algorithm is not robust.(2)The results of Quantum Genetic Algorithm experiment showed that if the test information on the cutoff score is the most important indicators, regardless of the flatness and time, can choose the population size of 80, the number of iterations for 500. When considering the three indicators, the amount of information to be as large as possible, flatness is also large, and the time is short we can choose the population size of 80, the number of iterations is 300.(3)After analysing the maximum test information on the cutoff score and information flatness near the cutoff score of GA and QGA by t test,the results showed that under the same population size and number of iterations.the maximum test information of GA is larger than QGA most of the time,but the amount of information flatness of QGA is significantly better than GA.In addition,the time of QGA is much short than GA, the robustness of QGA is greatly superior to GA,so the comprehensive performance of Quantum Genetic Algorithm is superior to the Genetic Algorithm in this item selection experiment.(4)Although the PSO algorithm was used to solve many other optimization problems,different c1 and c2 's value make the optimization results different. This paper used ANOVA to analyse whether different c1 and c2 would affect the test information on the cutoff score and information flatness near the cutoff score significantly.the result showed there is no significant difference,so c1 and c2 can be any value between [1, 4].(5)The results of Quantum Particle Swarm Optimization Algorithm for this item seleetion experiment showed that the inertia weight w1 and w2 had no significant influence to the maximum test information function, but had a significant effect on amount of flatness,so we must consider its value.The optimal parameter combination of QPSO are as following:if test information and flatness are the most important indicators, the best parameter combination is w1 is 1.2, w2 is 0.3, the number of the particle is 40 , the number of iterations is700.;When considering the three indicators,the best parameter combination is w1 is 1.2,w2 is 0.3, the particle number is 40, the number of iterations is 300.(6)After analysing the maximum test information on the cutoff score and information flatness near the cutoff score of PSO and QPSO by t test,the results showed that the information flatness of QPSO is not significantly better than PSO ,but the maximum test information of Q PSO is significantly larger than PSO .In addition,the time of QPSO is short than PSO, the robustness of QPSO is greatly superior to PSO,so the comprehensive performance of QPSO is superior to the PSO in this item selection experiment.(7)The results of Ant Colony Algorithm for this item selection experiment showed that if test information and flatness are the most important indicators,the optimal parameter combination is?=0.1,Q=150,m =70, d=360;when considering the three indicators, the optimal parameter combination is p = 0.5, Q = 250, m = 50,d= 200.(8)After analysing the maximum test information on the cutoff score and information flatness near the cutoff score of ACA and QACA by t test,the results showed that under nine parameters combinations,the maximum test information on the cutoff score of QACA is significantly larger than ACA's.the information flatness of two algorithm had no significant difference.in terms of the robustness of the algorithm,the results showed that in most combinations,QACA 's robustness is superior to the ACA.,and the time of QACA is significantly shorter than ACA.Therefore,considering all aspects of the evaluative indicators,Quantum Ant Colony Algorithm is superior to Ant Colony Algorithm in this item selection experiment.(9)Compared Quantum Genetic Algorithm,Quantum Particle Swarm Optimization algorithm,and Quantum Ant Colony Algorithm in two ways.the results of both methods showed that the Quantum Genetic Algorithm is the best in most of the evaluation indexes,especially the time of item selection is much short than other algorithms.All in all,Quantum Genetic Algorithm is the best algorithm in this research.