A New Combining Prediction Method of Visitor Numbers at Shanghai Expo

Abstract

Forecast of visitor numbers to the large-scale activities is the key issue of collective behaviors analysis and control. At present, forecasting visitor numbers is mainly based on traditional research approach or sole artificial neural network technology. Recent study results show that combining forecast model approach enjoys more precise forecast than monomial forecast approach. In this paper, a new forecast approach based on inflexion point was proposed. Then, we combined BP neural network and the inflexion approach to make comprehensive analysis and to predict visitor numbers to Shanghai Expo per day. Experimental results indicate that the proposed combining approach is feasible and effective in forecast of the visitor numbers, and is more precise in terms of monomial forecast method. Respectively, the average relative error of combining model is 0.1085, 0.1177, 0.1875 less than that of “inflexion” model, BP model and ARIMA model.