Real-Time Monitoring and Forecast of Active Population Density Using Mobile Phone Data

Abstract

Real-time monitoring and forecast of large scale active population density is of great significance as it can warn and prevent possible public safety accident caused by abnormal population aggregation. Active population is defined as the number of people with their mobile phone powered on. Recently, an unfortunate deadly stampede occurred in Shanghai on December 31th 2014 causing the death of 39 people. We hope that our research can help avoid similar unfortunate accident from happening. In this paper we propose a method for active population density real-time monitoring and forecasting based on data from mobile network operators. Our method is based solely on mobile network operators existing infrastructure and barely requires extra investment, and mobile devices play a very limited role in the process of population locating. Four series forecasting methods, namely Simple Exponential Smoothing (SES), Double exponential smoothing (DES), Triple exponential smoothing (TES) and Autoregressive integrated moving average (ARIMA) are used in our experiments. Our experimental results suggest that we can achieve good forecast result for 135 min in future.