Main Article Content
Abstract
Software development always faces a challenge in formulating mechanism for finding faults and developing effective methods and measures to remove these faults. Lowering the amount of defects from a software is a primary goal of software quality control. Software metrics and defect data is available to predict faults in software. This prediction helps in improving the software quality, proper resource allocation and reducing the testing effort. But sometimes software metrics and defect data is highly skewed towards the distribution of the non-faulty modules i.e. faulty modules are less in number than non-faulty modules. This is a class imbalance problem and data is imbalanced data. This study investigates the Random Forest algorithm for proposing software fault forecast model that deals with datasets that are not balanced. We used the dataset from NASA promise repository. PC5 dataset is used in our study to investigate Random forest. The result demonstrates that the accuracy of the proposed fault prediction model based on Random Forest is good and overall misclassification error is also less than the model based on Naïve Bayes i.e. the proposed model outperforms model based on Naïve Bayes.