Realtime Congestion Forecasting of Remote Space Through BLE Beacons

In this paper, we propose a system which forecasts the degree of congestions in a given space without actually visiting there. The proposed system is based on an assumption such that the arrival and departure of users concerned with the target space follows a specific probability distribution such as Gaussian mixture distribution and Poisson distribution. The system estimates parameters of the underlying probability distribution from time-series data reflecting the movement of users, and forecasts the degree of congestions at a certain time in the near future by using estimated parameters. The experimental results based on actual data acquired in a classroom of university show that the accuracy of parameter estimation could be comparable to that for complete data by filling missing future part with dummy data generated according to an appropriate normal distribution.