Web Service Anti-patterns Detection Using CNN with Varying Sequence Padding Size

Today, many IT companies use service-oriented architecture (SOA) as an effective architectural method for building their systems. Like many other complex structures, a service-based system (SBS) can be modified to meet new user needs. Continuous improvement of service-based systems to meet user needs will reduce the quality of software development leading to problems called Anti-patterns in web services. An anti-pattern is generally chosen as a good solution to the problem posed; however, it brings more liabilities than benefits. Anti-patterns inhibit the sustainability and perception of software systems. Therefore, there is an increasing need for the prediction of anti-patterns in the early stages of software design to enhance software quality in terms of maintenance costs, execution costs and memory consumption. In this paper, we analyse the effectiveness of convolutional neural network (CNN) with distinct padding sizes in the detection of web service anti-patterns. We conclude that CNN with two hidden layers is performing better with a mean accuracy of 97.58. SMOTE performs best among the data sampling techniques. We also inferred that the word embedding technique with a sequence padding size as 200 performed best with a mean accuracy of 97.76%.