Avoid Call Rejection in Overloaded Telecom Online Charging System for Priority Subscribers using Artificial Intelligence

Telecom prepaid services implement real-time charging with the help of the Online Charging System (OCS). OCS empowers service providers to charge a subscriber for services in real-time. OCS also authorizes the subscriber and validates if the subscriber has sufficient balance to allow the services. If allowed, it does rating and charges for the services used. In this article, we have discussed and proposed a solution for the problem faced when OCS is overloaded and starts rejecting new calls. This issue generally comes during the festive time or specific occasion when most of the subscribers want to use telecom services at the same time. The conventional charging system starts rejecting some pre-configured percentage of calls to overcome the OCS overload scenario. This may result in the rejection of some important calls, leading to a bad customer experience. The proposed solution given in this article suggests how to predict a priority subscriber, based on their payment history and usage of telecom services using machine learning, and to benefit these subscribers by allowing them to continue services even when OCS is overloaded. The proposed solution suggests using a classification model like XG Boost and Clustering algorithms like Gaussian Mixture Models (GMM) on different subscriber datasets and then combining the results obtained to predict priority subscribers. Such priority subscribers are allowed to continue calls in offline mode and are charged later using network events.