Design And Development Of Computer Adaptive Genetic Algorithm Test System

In recent years, the rapid development of the computer network has changed the mode of education. The learning mode of independent learning, collaborative learning based on constructivist learning theory and student-centered replaced the traditional teacher-centered teaching mode, and broke the limitations of time and space. At the same time, the traditional exam form based on pen and paper cannot meet the needs of today’s diverse educational, but computer-based adaptive testing (Computer Adaptive Test) is a new breakthrough to resolve this problem. The form of this examination not only reduces the duplication of the teachers and improves work efficiency, but also increases the papers’ objective and scientific nature, improves the quality of the examination, and realizes the measures from person to person.Today, the field of educational measurement has been widely introduced computer as an auxiliary means of measurement, such as a variety of computer level exam, English level exam or title examination, in which all the computer are taken as a testing tool, but the most of which change the manner from manual to computer. It is only the "paperless" form of traditional exam, and hasn’t a substantive breakthrough in the traditional classical test theory, so that "personalized test" cannot be realized. Therefore, this is not ideal education evaluation to adapt to the current teaching model.Based on item response theory IRT, this paper introduces the basic characteristics of the IRT, and the work principle of computer adaptive test based on the IRT, and summarizes the method of choosing starting point, proposes the improvement program of "two-step normalization". Through improvements the testing process, the test project which is far away the subjects ability is greatly reduced, the testing time and the calculation times are shorten, and at the same time the ability value is able to accurately be estimated. Combining adaptive test and test paper algorithm, based on computer adaptive test results, and using genetic algorithm for the examinee "tailor" to generate test paper, the paper realizes "measured from person to person". Based on genetic algorithm, the paper has developed an examination system with adaptive features and improved the genetic algorithm by artificial intervention. This method combined genes which were local optimal for individual and formed the next generation, which improve the efficiency of the genetic algorithm for test paper, is easy to operate, and can accurate positioning the level of the students, also can give different papers according to the different levels of different students for realizing the measures from person to person. Finally, the paper has finished the testing of the system, has carried on the analysis to the feasibility of the system, and statistically analyzed the experimental data to guide the work of the future system improvements.