A personalized genetic algorithm with forgetting factor for intelligent test generation

With the development of computer science and multimedia technology, computer-based testing becomes increasingly popular, especially the intelligent test generation systems. The algorithm used for generating a test paper has a direct impact on the quality and efficiency of intelligent test generating systems. Due to the advantages of parallelism and global space search, the genetic algorithms are recommended for solving the problem of an intelligent test paper composition. However, the traditional genetic algorithm has its own shortcomings, for example, it cannot create a personalized test paper for an individual learner, and it establishes a premature and slow convergence. This paper concerns itself with each user's current knowledge level and the extent in which a learner forgets. It keeps to the basic principles of Psychology inasmuch as those principles relate to memory and natural memory loss. Utilizing the genetic algorithm, a Personalized Genetic Algorithm with Forgetting Factor (PGAFF) is proposed and used for a multi-constrained test paper composition problem. Experimental results show that the proposed algorithm can support the personalized test generation which can select questions that users haven't mastered well to composite a test paper. The generated test can help testers find out those questions that they don't know well and those they may have forgotten. In this view of point, we can see that PGAFF outperforms existing simple genetic algorithm on the intelligence of test generation.