A Novel Software Fault Prediction Approach To Predict Error-type Proneness in the Java Programs Using Stream X-Machine and Machine Learning

Software fault prediction makes software quality assurance process more efficient and economic. Most of the works related to software fault prediction have mainly focused on classifying software modules as faulty or not, which does not produce sufficient information for developers and testers. In this paper, we explore a novel approach using a streamlined process linking Stream X-Machine and machine learning techniques to predict if software modules are prone to having a particular type of runtime error in Java programs. In particular, Stream X-Machine is used to model and generate test cases for different types of Java runtime errors, which will be employed to extract error-type data from the source codes. This data is subsequently added to the collected software metrics to form new training data sets. We then explore the capabilities of three machine learning techniques (Support Vector Machine, Decision Tree, and Multi-layer Perceptron) for error-type proneness prediction. The experimental results showed that the new data sets could significantly improve the performances of machine learning models in terms of predicting error-type proneness.