An Extended Assessment of Data-Driven Bayesian Networks in Software Effort Prediction

Software prediction unveils itself as a difficult but important task which can aid the manager on decision making, possibly allowing for time and resources sparing, achieving higher software quality among other benefits. Bayesian Networks are one of the machine learning techniques proposed to perform this task. However, the data pre-processing procedures related to their application remain scarcely investigated in this field. In this context, this study extends a previously published paper, benchmarking data-driven Bayesian Networks against mean and median baseline models and also against ordinary least squares regression with a logarithmic transformation across three public datasets. The results were obtained through a 10-fold cross validation procedure and measured by five accuracy metrics. Some current limitations of Bayesian Networks are highlighted and possible improvements are discussed. Furthermore, we assess the effectiveness of some pre-processing procedures and bring forward some guidelines on the exploration of data prior to Bayesian Networks´ model learning. These guidelines can be useful to any Bayesian Networks that use data for model learning. Finally, this study also confirms the potential benefits of feature selection in software effort prediction.