Optimizing Prediction Time and Accuracy of Users’ Regular/Irregular Grouping Using ML Classification

Utilization of Artificial Intelligence in mobile networks has seen significant growth in the last decade. It spans from mobile phone applications to various aspects of mobile network operations, including planning and optimization. Understanding and modeling users’ mobility patterns enhance mobility-based operations such as handover, scheduling, paging, and caching. In the literature, numerous models exist for grouping users’ mobility patterns in mobile networks based on their regularity. In our paper, we employ machine learning techniques to achieve this task, prioritizing low prediction time and high accuracy. The dataset used in our work is provided by an Internet Service Provider in Shanghai city, covering a span of six months. User records represent the sequence of serving base stations over four consecutive weeks. We employ various Machine Learning classification algorithms. The results demonstrate that our proposed coding method and the procedure of unifying the number of records, followed by the use of the Random Forest algorithm, yield the best results.