Mining of Software Patterns using Object Oriented Metrics and Machine Learning Techniques

Development of desired software in the present day scenario is becoming too much complex corresponding to the ever-growing complex user requirements. Hence there is a need for developing the right methodology for solving these complex problems. In order to address these issues in design phase, a number of tools and techniques are available and one of them is the use of design patterns. These patterns help to find a better solution for the problems, which are recurring in nature. It is often desired to detect design patterns from the source code of similar category of softwares, as it improves maintainability of the source code. In this study, mining of design pattern technique has been proposed, which is based on supervised learning techniques and object-oriented software metrics. During the pattern mining process, object-oriented metrics-based dataset is prepared. Subsequently, machine learning techniques such as Artificial Neural Network and Random Forest are applied for the pattern mining process. For the critical examination of the proposed study, data from an open source software e.g., JUnit and JHotDraw are considered for the mining of software design patterns