Software Defect Prediction using Oversampling Algorithm: A-SUWO
Shabrina Choirunnisa, Biandina Meidyani, Siti Rochimah · 2018
To predict software defects required prediction models using defect data and software metrics called Software Defect Prediction (SDP). Some learning algorithms are used to identify possible decay to program modules, thus affecting the optimum utilization and allocation of resources. However, the accuracy of classification is influenced by the robustness and quality of data. The class imbalance in the data will affect the accuracy of predicting defect or not defect. To improve the accuracy of the Software Defect Prediction (SDP) model, we propose a new framework using A-SUWO to handle the class imbalance. Data with a balanced class will be classified to produce an accurate class. The dataset used is NASA. Using A-SUWO shows that the proposed framework can predict defects effectively. The highest accuracy is shown by A-SUWO-Random Forest.