Using product, process, and execution metrics to predict fault-prone software modules with classification trees

Software-quality classification models can make predictions to guide improvement efforts to those modules that need it the most. Based on software metrics, a model can predict which modules will be considered fault-prone, or not. We consider a module fault-prone if any faults were discovered by customers. Useful predictions are contingent on the availability of candidate predictors that are actually related to faults discovered by customers. With a diverse set of candidate predictors in hand, classification-tree modeling is a robust technique for building such software quality models. This paper presents an empirical case study of four releases of a very large telecommunications system. The case study used the regression-tree algorithm in the S-Plus package and then applied our general decision rule to classify modules. Results showed that in addition to product metrics, process metrics and execution metrics were significant predictors of faults discovered by customers