A Machine Learning Based Positioning Scheme Using Cellular Network Data

The evolution of the fifth generation (5G) new radio (NR) is geared towards offering enhanced flexibility to accommodate evolving service demands. On the other hand, machine learning (ML) has showcased its effectiveness across a diverse range of tasks, such as pattern recognition, algorithmic trading, content generating, and natural language processing. Notably, ML exhibits performance scalability in tandem with the volume of accessible data. However, in NR, the precision in ascertaining user locations remains a critical challenge for mobile operators during the strategic planning and fine-tuning stages of cellular networks. Creating a methodology for mobile users' positioning can enhance resource management efficiency, leading to greater economic benefits. In this paper, we present a ML based mobile users' positioning scheme relying on support vector machine (SVM) and bayesian learning. Experiment results indicate that the average positioning deviation is about 0.5 meters with 80% confidence level in real-world operating field.