| With the development of computer based education and artificial intelligence,more and more experts focus their attention to the research on the system of auto-generating test paper.Some scholars have designed a wide range of auto-generating test paper system at present,but these systems are with more or less shortcomings.Some systems are designed so dependent on the item bank that they can not be commonly used. Some systems' questions attributes are too simple.Although they work fast,these systems cannot offer objective and scientific evaluation of the quality of the test questions with their inadequate reflect of the properties of test paper.These systems usually have fast speed,but the evaluation of system quality are not objective,scientific because of the reflect the quality of the properties of questions are not enough.Some systems are designed too complicated in order to pursue high quality of test paper,but the price is low efficiency.On the basis of analyzing a great deal of articles at home and abroad,this thesis probes the theoretical foundation of the item bank construction,and analyzes the testing standard of the test paper,as well as the parameters of the test papers, then creates the test paper' s mathematics model on the constraints of types of question,knowledge points,degree of difficulty,degree of distinguish,score,answer time, exposition,the deadline etc.Secondly,this thesis selects genetic algorithm as the way to produce test paper based on the actual problem,and selects some kinds of algorithms as improved genetic algorithm on the basis of in-depth study of genetic algorithm.Thirdly,the thesis puts forward a strategy of generating test paper while establishing the item bank,and gives an improved test paper auto-generating process based on genetic algorithm.At last,it also designs the software,and carries out a large number of experiment.The experimental results shows that the algorithm is efficient,reasonable and effective.
|
PDF Full Text Request