Prediction of Web Service Performance and Successability using Comparative Analysis of Machine Learning and Deep Learning Algorithms

Abstract

Internet services, also known as e-services, has gained in importance as a means of providing online commercial services. Service Oriented Architecture (SOA) is built on a combination of multiple web services, each responsible for developing a specific task, in order to obtain complete professional software. Quality of Web Services (QWS) is a key characteristic for choosing a web service throughout the service configuration procedure and has a set of non-functional properties such as response time, availability, throughput, successabilty, reliability, compliance, best practices, latency and documentation is included. Now a days, Machine Learning (ML) has been used for service classification and regression problems. Though, the performance of traditional ML techniques is highly dependent on the superiority of physical feature engineering. We propose a technique to extract multiple data, procedural, and structured set metrics from a web service interface and use them as predictors to estimate QWS properties. Our proposed method applies Deep Learning (DL) methods with six dissimilar training approaches to build predictive models with successabilty rate. The outcome of the research shows that the proposed method is efficient and the investigational outcomes indicates that operational quality metrics are superior to technical and data quality metrics in terms of Mean Absolute Error (MAE), Mean Square Error (MSE), Root Mean Square Error (RMSE), Root Mean Square Logarithmic Error (RMSLE), R-Squared and Adjusted R-Squared performance metrics. The comparative study of six models concludes that Extra Trees Regressor model outperforms other five common DL training methods.