The Research And Application Of Massive Mobile Location Data

In recent years,with the rapid growth of population,Road congestion,Scenic crowded usually happened when people enjoying holidays,Organizing large-scale activities or holding competitions.And even stampede accident turned out,causing many casualties.The most tragic accident happened in Mecca,1990,1426 pilgrims was trampled or suffocated to death.For an another example,36 people died and 47 people wounded in the stampede accident in Shanghai,2014.The occurrence of stampede accidents is mainly because the lack of effective monitoring and forecasting from local government departments,scenic management departments.This makes the scenic capacity overload severely and occurs the vicious accidents.This paper will research for a regional crowd flow monitoring system that helps relevant departments to estimate,monitor the crowd flow in a region,and alarm in real-time.It will guide related staffs,traffic police to evacuate crowd,reduce the scenic spots load and finally avoid the stampede accident.Flow forecast method is an important area in flow monitoring system.In this paper,we use history data to estimate the crowd flow by the time series algorithm model.And then filter the characteristic of crowd flow from the massive data in mobile base stations to make the system more accurate.After getting valid data from massive data of position moving,we conducted the following research and application with Hadoop,Kafka and some other software:1.Position data is the key point.Because of the redundancy and hugeness of the data that gets from the base station directly,preprocessing for this data is very important.In our company’s data collection platform,we use regular expressions as the filter conditions.To resolve the problem of input / output data synchronization,we use Kafka message middleware.2.The traditional database is not that high-performance to store massive position data.So we use Redis and HBase to store the filtered data.3.We use time series prediction algorithm model in the core of crowd flow estimating.By calculating coefficient ratio of the preceding term and then calculating weighted average with current data,we can estimate crowed data in next time period.Simultaneously,some other data will be used to build a simple model of crowed flow characteristic that helps system run better.