Research On Fast Calculation Method Of Disaster-affected Population Based On Mobile Phone Signaling Data

In recent years,natural disasters and occasional emergencies have become more frequent in China,bringing huge losses to the country and people.As the main carrier after the disaster occurs,how to quickly obtain the affected population distribution and accurately assess the number of the affected population after the disaster occurs is the primary task to reduce casualties,timely carry out emergency rescue,and is also a reasonable allocation of relief materials,disaster relief personnel premise.However,the construction of China’s emergency surveying and mapping system is still in the initial stage,mainly focusing on the acquisition of basic data,such as the disaster area image,3d model,etc.,based on the basic data of the disaster area can determine the scope and extent of the disaster,but the population distribution in the disaster area can not be timely and effectively acquired.In order to solve the problem that emergency mapping cannot effectively assess population disaster,this paper finds through reading domestic and foreign literature that there are two main types of population spatial and temporal distribution models: population spatial distribution based on grid data and population spatial distribution based on spatial and temporal big data.In general,the former is based on the population data obtained by the unit of administrative boundaries and USES some mathematical model or algorithm to disperse the population into the grid.The disadvantage of this method is that the spatial and temporal resolution of the population data is too large to reflect the real-time distribution of the population.Although the population spatial distribution method based on spatial and temporal big data can reflect the population distribution in real time,the spatial and temporal big data is usually composed of a series of discrete points with coordinate information,which is not conducive to the study of the law of continuous population distribution.Therefore,based on the two methods,this paper proposes a population dynamic distribution based on cellular signaling data.Paper first use of distributed computing framework for mobile signaling data in datacleaning,cleaning content includes: the lack of data,repetition,error,because the data after cleaning has "more than one machine" phenomenon,need to repeat the user data for secondary cleaning,clean methods using ST-DBSCAN clustering algorithm to extract the user travel path,when multiple track overlap,only keep a track data,delete the duplicate data,mobile phone signal data at this time to the representative of the real user information,and then after cleaning the phone signaling data stack to the study area boundary in the grid,and by using time series prediction model: The differential autoregressive moving average model(ARIMA)is used to predict grid data.When the time series data do not conform to the randomness test of ARIMA model,the moving average method is used to predict grid data.The prediction error is evaluated by using the absolute error method,and the error frequency in different ranges is calculated.Finally,the model is integrated into the emergency surveying and mapping system.By using the research method in this paper,more accurate population distribution and population number statistics can be obtained in the first time.At present,this method should be integrated into the emergency mapping system...