Impact of Parameter Optimization on Heterogeneous Binary Classifiers for Software Defect Prediction Frameworks

Abstract

Machine learning classifiers consist of a set of parameters. The efficiency of classifiers in the context of software defect prediction is greatly impacted by the parameters chosen to execute the classifiers. Mostly, the researchers rely on default parameters. However, these parameters can be configured to achieve more accurate results. Blockchain technology can also play a role in software defect prediction by providing a secure and transparent environment for storing and analyzing data related to software development and testing processes. This can help to prevent data tampering and ensure the integrity of the information utilized to train machine learning models for defect prediction. In this research, the efficiency of machine learning classification algorithms for software defect prediction is analyzed through parameter optimization. Three heterogeneous binary classifiers i.e. Decision tree, Support vector machine, and Naïve Bayes are selected to examine the results of parameter optimization. The experiments are performed on seven publicly available NASA Datasets. These datasets contain historical data from NASA software. The dataset is split into 70-30 proportions with class preservation. To evaluate the performance; five statistical measures have been implemented i.e. precision, recall, F-Measure, the area under the curve (AUC), and accuracy. The results of the research reveal that there is a clear distinction in the results achieved by executing binary classifiers with default parameters and with optimized parameters; hence the performance of binary classifiers in the context of software prediction can be enhanced to a great extent by employing parameter optimization.