Research on intelligent test paper auto-generating algorithm based on improved GA

A high quality test paper auto-generating is generally at the cost of considerable time. In order to handle the contradiction between quality and speed of combination, an improved subsection-single-point crossover and mutation GA based on real segment coding and condition matrix is proposed in the thesis. In the design of genetic operators roulette wheel selection, subsection-single-point crossover and mutation policy are synthetically used. And the population can be evolved continuously by the optimal-store policy. The algorithm has been proved feasible and effective through test and comparison between other algorithms. At last the validating system enhances the users´ condition constraints for test paper through manual inching, thus the system is more simple and practical.