Multiple phases-based classifications for cloud services

Abstract

The current problem in cloud services discovery is the lack of standardisation in the naming convention and the heterogeneous type of its features. Therefore, to accurately retrieve the appropriate services, an intelligent service discovery is required. To do that, the cloud services attributes should be extracted from the heterogeneous formats and represented it in a uniform manner such as ontology to increase the accuracy of discovery. The extraction process can be done by classifying the cloud services into different types. In this paper, single and multiple phases-based classifications are performed using support vector machine (SVM) and naïve Bayes as classifiers. The Cloud Armor's dataset used which represents four classes of cloud services. Topic modelling using MALLET tool is used for dataset pre-processing. The experimental results showed that the classification accuracy for the two phases-based and single phase-based classifications reached 87.90% and 92.78% respectively.