Heterogeneous Defect Prediction Using Ensemble Learning Technique

Abstract

One of the quite frequently used approaches that programmers adhere to during the testing phase is the software defect prediction of the life cycle of the software development, this testing becomes utmost important as it identifies potential error before the product is delivered to the clients or released in the market. Our primary concern is to forecast the errors by using an advanced heterogeneous defect prediction model based on ensemble learning technique which incorporates precisely eleven classifiers. Our approach focuses on the inculcation of supervised machine learning algorithms which paves the way in predicting the defect proneness of the software modules. This approach has been applied on historical metrics dataset of various projects of NASA, AEEEM and ReLink. The dataset has been taken from the PROMISE repository. The assessment of the models is done by using the area under the curve, recall, precision and F-measure. The results obtained are then compared to the methods that exist for predicting the faults.