| Being popular for the network education, the problem of generating test paper intelligently has been widespread concern. The problem of generating test paper intelligently is actually a multi-objective optimization problem. The traditional method in generating test paper, such as the random selection and back tracking, has obvious defects on the time and the success rate of generating test paper. Therefore it is difficult to find an effective solution. According to the old research, how to find a useful method to solve the problem, which can aim at the actual characteristics of generating test paper intelligently, has become a hot spot of concern to everyone.In this article, based on Computer Rand Examination in Software Institute of Dalian Jiaotong University, we research on how to combine algorithm of generating test paper with the existing item bank and how to obtain questions which satisfy the user's request from item bank, so that this test paper is able to objectively reflect the real capacity level of the user. According to the research of algorithm and basic principle of generating test paper, we develop one module of generating test paper intelligently, which is based on Improved Genetic Algorithm. According to difficulty supplied by the user and scores of test paper, which are important attributes of questions, we can get dynamic generating test paper.Through analysis and comparison with the genetic algorithm and common algorithm of generating test paper intelligently, in this article, a new method of generating test paper intelligently based on the improved genetic algorithm is given. This method uses the partition real number coding, it can get over some shortcomings of traditional binary coding, such as searching space is too large and the length of coding is too long. This method also can save the time of decoding and improve the efficiency of the algorithm operation. In this paper, there is a new method of mutation rate computation. To ensure individual ability of evolution and effectively prevent the algorithm into local optimal solution, this article introduces a new coefficient km which changes exponentially and makes mutation rate reduce along with the evolving algebra. For the characteristic of partition real number coding, in the process of mutation operation, we use basic mutation operation in the section, then rant a sequence whose length is equivalent to coding length. Through result of the comparison between each number in sequence and mutation rate, we judge this pot execute mutation or not. This method improved the capability of local searching of the genetic algorithm, maintained the multiplicity of community, and ensured the convergence of the genetic algorithm.This algorithm is applied to the actual online examination system. Both the theoretical analysis and the experimental comparison show that improved genetic algorithm can satisfy the needs for actual examinations. Improved genetic algorithm has more ability to get over premature convergence than the traditional algorithms. Moreover, improve genetic algorithm can obviously improve the ability of global research and the convergence speed than tradition genetic algorithm. It proves that the improved algorithm is effective and superior. |
