Abstract
With the rapid development of web service technologies, the number and variety of web services available on the internet are rapidly increasing. Currently, service registries support human classification, which has been observed to have certain limitations, such as poor query results with low precision and recall rates. With the huge amount of available web services, efficient web service discovery has become a challenging issue. Therefore, to support the effective application of web services, automatic web service classification is required. In recent years, many researchers have approached web service classification problems by applying machine learning methods to automatically classify web services. The ultimate goal of our work is to construct a classifier model that can accurately classify previously unseen web services into the proper categories. This paper presents an intensive investigation on the impact of incorporating feature selection methods (filter and wrapper) on the performance of four state-of-the-art machine learning classifiers. The purpose of employing feature selection is to find a subset of features that maximizes classification accuracy and improves the speed of traditional machine learning classifiers. The effectiveness of the proposed classification method has been evaluated through comprehensive experiments on real-world web service datasets. The results demonstrated that our approach outperforms other state-of-the-art methods.