Application of improved genetic algorithm in automatic test paper generation

Abstract

Automatic test paper generation system is to automatically generate papers by computer from test database with many constraint conditions according to requirements of teachers and teaching. It could greatly reduce teachers' work, and make the difficulty coefficient of test paper reasonable. The system plays an important role in reform of examination system. Traditional genetic algorithm uses binary code, but because the binary string is too long, it cannot control well the number of question types. However, the system with decimal code could avoid that problem. In the process of genetic manipulation, crossover operation takes subsection crossover, i.e. single point crossover within one question type. Therefore, the whole chromosome is multi-point crossover, which makes the result more reasonable. This system makes use of global optimization and fast convergence speed of genetic algorithm to design an intelligent algorithm for automatically generating test papers. We have established and described the chromosome structure of test paper and the fitness function, designed genetic operators, and completed corresponding genetic algorithm application software to realize the automatic generation of test papers. Experimental results show that the automatic test paper generation system based on genetic algorithm achieves optimization of efficiency and reasonability of difficulty coefficient of test paper.