城市尺度下基站人群时空预测模型

Spatio-temporal analysis and forecast of population of cellular in city scale

利用3G用户上网数据推演了群体分布动态聚散过程,并依此提出了基站人群时空预测模型与方法。相比于以往单从时间序列学习或从整体空间时空学习预测的方法,避免了对时空信息平滑作用的影响。经实验验证,该预测模型在细粒度的人群预测上有更好的预测性能,尤其适用于对突发性人群的预测,有助于更好地理解人群分布,并为移动网络优化管理提供很好的理论指导。

This paper studied the time and space distributions of city-scaled residents by making use of 3 G mobile network da- ta, and proposed a spatio-temporal forecasting model of cellular population. The inspiration came from the real world accumulation and dispersion of crowd. For importing the dynamic mobility, this model could avoid the effect of data smoothing widely embedded in previous models based on time series and overall space data. Through experimental verification, this model has been proved that it owns a better prediction performance on a fine-grained scale, especially in the prediction of incident crowd. This study and mode helps to better understand the crowd distribution and provide support for wireless network base station research.