Mobile Service Experience Prediction Using Machine Learning Methods


Yigit I. O. , Ciftci S., Kalyoncu F. A. , Kaya T.

26th IEEE Signal Processing and Communications Applications Conference (SIU), İzmir, Türkiye, 2 - 05 Mayıs 2018 identifier identifier

Özet

With the introduction of 4.5G, mobile operators have focused their efforts, infrastructure investments, tariffs and advertisements on the improvement of mobile data rates and services. Mobile services provided by mobile operators are influenced by various factors like the regional coverage of the operator, usage traffic, time and weather conditions. As a result, there may be differences between the quality of mobile services that the operators offer to their customers and those that the customers can actually access. The purpose of this study is to suggest a modelling approach for the prediction of the mobile service types that customers can experience based on machine learning techniques. To do this, based on 2017 speed tests data of three operators, alternative classification models are constructed for the prediction of the mobile service type. By comparing the performances of the models, best classification models were determined for different service categories. Using the data obtained from mobile speed tests performed on a limited number of locations, the models developed here enable the prediction of the possible service types that customers can experience in all locations in which the operators serve.