The Impact of Instance Selection Algorithms on Maintenance Effort Estimation for Open-Source Software

Open-source software are very used nowadays in the industry, and the performance of the estimation of their maintenance effort becomes an interesting research topic. In this context, researchers have conducted many open-source software maintenance effort estimation (0-MEE) studies based on statistical and machine learning (ML) techniques for better estimation. This study focuses on the impact of instance selection on the performance of ML techniques in 0-MEE, mainly for bug resolution. An empirical study was conducted using three techniques: K-nearest neighbor (kNN), support vector machine (SVM), and multinomial naive Bayes (MNB) using all-kNN instance selection algorithms on three datasets: Eclipse JDT, Eclipse Platform, and Mozilla Thunderbird datasets. This study reports on a set of 18 experiments and a comparison of the results. The results of this study show that instance selection helped make ML techniques more performant.