Study On The Prediction Of Regional Stream Of People Density Based On Ensemble Learning

The occurrence of these tragedies is not only because China has a vast territory,large population,with the people's standard of living is getting better and better,people are increasingly high demand for spiritual life,most especially young people are more willing to to the community to participate in a variety of interesting and fun activities,however due to various reasons,in some public places often appear crowded situation,which not only affects the tourists travel mood more visitors to bring a great security risk.Although the flow of people through the relevant measures to manage the flow of people,to a certain extent,to ease the congestion situation,but in the open area,this approach is no longer applicable.In this paper,the public area of the flow prediction technology can provide a strong basis for management,has important practical significance.This study is based on the Storm platform,the collected data are processed to retain the density of people in a certain period of time.The prediction model was established by the existing historical data,namely,the density of the stream of people,and the density of the flow at the specified time was predicted.The prediction algorithm in this paper involves mainly including ARIMA,GM and grey prediction RBF neural network three.First,according to the characteristics of different flow density data characteristics and ARIMA model,GM two prediction algorithm alone on different data model was established,and the and the algorithm of integration through the results of two predictive models are integrated to get GM-ARIMA predictions;and then,the prediction error of GM-ARIMA using RBF neural network model,to predict GM-ARIMA model of error;finally,to GM-ARIMA forecast data and RBF prediction error data were combined to form the final prediction.In the choice of ARIMA and GM algorithm based on the characteristics of the basic model and algorithm flow density data reference to the optimization model proposed by other scholars,according to the minimum sum of square error principle,according to certain rules of two kinds of prediction results are integrated,so as to establish prediction model based on ARIMA and grey model of regional flow density.In this paper,a detailed theoretical method and design procedure are presented,and the prediction scheme is carried out on the Storm distributed platform.Experimental results show that GM-ARIMA integrated prediction effect than individualGM or Arima forecast effect is good,and after correction residual prediction effect is better than the separate effects of integration,according to the mean square error calculation,GM-ARIMA than GM improve the 56.05%,4.5% higher than that of ARIMA,residual modification results than GM-ARIMA improve 77.1%.Therefore,the study of flow density measurement on to make a positive contribution.