TLHEL: Two Layer Heterogeneous Ensemble Learning for Prediction of Software Faults

摘要

Software fault prediction is the most ubiquitous research concept in the domain of software engineering. Prior literature concedes the importance of ML techniques in the prediction of software faults, but the expedient method that gives consistently good results has still remained undetermined. So, to achieve high accuracy consistently, we have investigated many ensemble methods that advance the individual techniques and improves the performance of the fault prediction model. This paper proposed the novel TLHEL twolayer heterogeneous ensemble model to predict software faults with less misclassification rate. The novelty of this model is that it combines the metric selection and training as a single process which reduces the computation overhead significantly, and performs feature selection with crossvalidation, which particularly reduces the biasness of the model. The implementation of the TLHEL model will significantly increase the efficiency of the model.