Abstract: In order to design effective software defect prediction model, this paper proposes an approach combining Particle Swarm Optimization(PSO) algorithm and Na?ve Bayes(NB). After discretizing the original data, the error rate of NB is taken as fitness function of the particle, and a software defect prediction model is constructed. It applies one software project JM1 data of NASA to implement the simulation experiment. The results show that the approach proposed has lower error rate than other methods, and has good performance.

Key words: software defect, prediction model, Particle Swarm Optimization(PSO), Na?ve Bayes(NB), data discretization

摘要： 为了设计高效的软件缺陷预测模型，提出一种将粒子群优化算法与朴素贝叶斯(NB)相结合的方法。该方法对历史数据进行离散化后，以NB分类的错误率作为粒子适应值函数，构建软件缺陷预测模型。通过对美国国家航天局软件工程项目的JM1数据进行仿真实验，证明该模型在预测性能方面优于同类方法，预测效果良好。

关键词: 软件缺陷, 预测模型, 粒子群优化, 朴素贝叶斯, 数据离散化

CLC Number: 

• TP311.5