Prediction of users charging time in cloud environment using machine learning

Due to the rapid growth of mobile devices, and the new generation of mobile technologies and services, terms such as ‘mobile charging’ and ‘billing processes’ arise and become a hot area for researchers and mobile telecommunications operators. Supporting the new revenue schemas and different pricing points requires a progressive improvement not only on the level of platform/applications or the physical infrastructure, but also on the level of cloud resources across different layers both IaaS and PaaS. Operators avail such cloud resources to vendors to manage the users charging transaction requires a certain performance management and enhancement. Analytical models and machine learning techniques are employed to manage, analyse the users’ data/logs and to get more useful info that can be used in the management of the cloud resources in order to reach the best resources utilization with the highest revenue stream from the services users. Machine learning techniques are used to predict users charging behaviour based on their previous charging history and they are grouped into a set of clusters based on the similarity of the charging logs in a self-adaptive model that learn from old and current charging transactions as well. A detailed experiment is conducted to show how to reduce the number of charging transactions to the minimum that does not affect the revenue stream and at the same time leads to the best resources utilization. Finer tuning on this is made by applying forecasting and prediction techniques on the data to enhance the result. Several prediction techniques are applied to reach the highest accuracy level of prediction. Numerical results serve to confirm the accuracy of the proposed analytical model while providing insight on how the different parameters and designs affect cloud resources performance.