摘要

In this paper, BS-GEP (Gene Expression Programming based on Block Strategy) is proposed as an improved GEP, in consideration of the characteristics of software reliability growth models. The block mechanism is the key point of BS-GEP, in which the population is divided into several blocks according to the individual fitness of each generation and the genetic operators are reset differently in each block to guarantee the population diversity. Based on BS-GEP a new software reliability modeling method is formed, and the final model is obtained by training on a real failure data set. Finally, the model is compared with several representative ones through case studies. The results show that the BS-GEP model provides better fitting and prediction performances.