# An Approach for Spatial-temporal Traffic Modeling in Mobile Cellular Networks

## TL;DR

This paper introduces a combined spatial-temporal traffic model for cellular networks, using sinusoid superposition for temporal variation and lognormal distribution for spatial distribution, validated with real data.

## Contribution

It presents a novel integrated model for traffic variation in cellular networks, capturing both temporal and spatial dynamics based on real data.

## Key findings

- The sinusoid superposition effectively models temporal traffic variation.
- Lognormal distribution accurately describes spatial traffic distribution.
- Models closely match real traffic data in different regions.

## Abstract

The volume and types of traffic data in mobile cellular networks have been increasing continuously. Meanwhile, traffic data change dynamically in several dimensions such as time and space. Thus, traffic modeling is essential for theoretical analysis and energy efficient design of future ultra-dense cellular networks. In this paper, the authors try to build a tractable and accurate model to describe the traffic variation pattern for a single base station in real cellular networks. Firstly a sinusoid superposition model is proposed for describing the temporal traffic variation of multiple base stations based on real data in a current cellular network. It shows that the mean traffic volume of many base stations in an area changes periodically and has three main frequency components. Then, lognormal distribution is verified for spatial modeling of real traffic data. The spatial traffic distributions at both spare time and busy time are analyzed. Moreover, the parameters of the model are presented in three typical regions: park, campus and central business district. Finally, an approach for combined spatial-temporal traffic modeling of single base station is proposed based on the temporal and spatial traffic distribution of multiple base stations. All the three models are evaluated through comparison with real data in current cellular networks. The results show that these models can accurately describe the variation pattern of real traffic data in cellular networks.