An Enhanced Bayesian Decision Tree Model for Defect Detection on Complex SDLC Defect Data

Abstract

In this paper, we explore the multi-defect prediction model on complex metric data using hybrid Bayesian network.Traditional software metrics are used to estimate the effect of defects for decision making. Extensive study has been carried out to find the defect patterns using one or two software phase metrics.However, the effect of traditional models is influenced by redundant and irrelevant features.Also, as the number of software metrics increases, then the relationship between the new metrics with the traditional metric becomes too complex for decision making. In this proposed work, a preprocessed based hybrid Bayesian network was implemented to handle large number of metrics for multi-defect decision patterns. Experimental results show that proposed model has high precision and low false positive rate compared to traditional Bayesian models.