A Comparative Analysis on the Detection of Web Service Anti-Patterns Using Various Metrics

Abstract

Nowadays, the application of machine learning for developing prediction models is one of the most critical research areas. Early prediction of anti-patterns using machine learning can help developers, and testers fix the design issues and utilize the resources effectively. This work analyzes four different sets of metrics, i.e., source code, WSDL, text, and sequence metrics, to develop web service anti-pattern prediction models. These sets of metrics are treated as an input for models trained using thirty-eight classification techniques to build a model. The experimental finding shows that the models trained using sequence metrics produce better results. The experimental finding also confirmed that the models trained on balanced data achieved better performance than the original data. Further, it is also found that the models trained using CNN and LSTM deep learning approach achieve better results compared to other techniques.