Abstract:

In most of the automatic test paper generation system, the constraint conditions are not perfect, and the results are not ideal. In view of this situation, we put up a new method called genetic algorithm with local search; it can solve the problem of constraint conditions which are not perfect, and can achieve the better results, after the teacher’s certification, it proves to be applied to the actual teaching. The algorithm uses the subsection encoding based on the type of questions. Three genetic operators adopt the following strategy: the subsection crossover strategy according to the type of the questions, which can ensure the capability of global search and the total number of selected questions in each type are not changed after this operation, the mutation strategy based on the tabu table local search, which can ensure the random and correlative search on the item pool, and improve the search ability of this algorithm, and the μ+λ strategy which is always used in combinatorial optimization evolutionary algorithm to enhance local search of this algorithm. The results show that in the same number of iterations, the optimal solution which can be found in the new algorithm is obviously better than that of tradition algorithm.

Key words: computer assistant instruction, genetic algorithm, combinatorial optimization, local search, automatic test paper generation