Analyzing Mobility Patterns for Location Prediction and Target Coverage.pdf

1、Analyzing mobility patterns has been a recurrent research topic in recent years.This is due to the fact that the end product of movement always has an effect either good or bad.Accurately using mobility patterns for effe

2、ctive applications has been a recurrent issue in modern mobility research.This is because platforms required to put the results of mobility analysis has not presented itself in a more clear terms until now.With the adven

3、t of technology and growing hunger of users for more information and quick rendering of services, mobility outcomes now has a somewhat use.Even with this trend,little has been done in the area of effective usage of these

4、 outcomes.Especially as seen in the area of location prediction and target coverage.
　　Past researches have been done in these regard, however, they have been limited either because of the methods used or inability to

5、 get accurate data for this.In this work, we propose the development of two algorithms to help put to use the outcomes of users mobility.
　　The first, Time dependent multiclass support vector machine (T-MSVM) involves

6、 using a particular time to predict the user's future location based on their locations at that time in the past.Thus, allowing prediction results to be more accurate since the next location of a user will be narrowed to

7、 a specific time frame.First, we develop the T-MSVM model, and then we build the proposed prediction algorithm based on this model.Through experiments, we show that the proposed T-MSVM algorithm can achieve an accuracy o

8、f 90％ over a week period and more than 95％ accuracy over a month period in predicting the next location of a user.
　　The second is an algorithm based on queuing theory in tandem with mobile crowd sensing which is for

9、target coverage.To do this, first, we develop some models which are based on the birth-and-death mechanism (one of the tools in queuing theory) to determine how long a target has to wait, the mean busy period of sensors

10、and mean idle period of sensors.While developing these models, we considered cases where there exist a single sensor and n-sensors in the system.Based on these models, we developed the required algorithm.The simulation r

11、esult shows that as the number of sensors increase relative to the number of targets, an average time before a target gets discovered is 0.2 seconds and sensor utilization decreasing towards zero as the number of sensors