Forecasting Model Based On Intelligent Computing Research And Its Application In Public Crisis Management

As a widely discussed issue, there are already a large number of models to solve forecasting problems. In this research, these models can be roughly divided into two categories:time series models and Spatial-temporal Analysis combined with GIS.The time series forecastiong model mainly contain linear and nonlinear ones. Linear models such as ARMA, ARIMA models can adjust itself according to the new obtained data so as to improve the forecasting accuracy, however, they are unable to analysis all the factors of fluctuations in the data, and thus the forecasting accuracy will be affected. The typical nonlinear model is artificial neural network model,which has a strong parallel computing capabilities. It can implement any complex nonlinear mapping function proved by mathematical theories. It has been widely applied in the field of forecasting, but it is can lead to local minima and over-fitting phenomenon. Chaos Particle Swarm Optimization is an evolutionary computation technique combined with chaos search, it depends on sharing the individual information with the whole group in the problem solving space in order to find the optimal solution and the movement of particles evoluted from disorder to order. So Swarm intelligence technology has been widely applied to optimize fields, which can effectively find the global optimal solution. This study proposed a Chaos Particle Swarm optimization BP neural network prediction model as a basis research model based on the two intelligent calculation methods. This study has paid more attention on these two aspect as follow:1) The parameters selection of the hybrid model. Through a large number of experiments in-depth analysis the performance of the hybrid model of neural network based on smart optimization algorithms, experiment results demonstrate that without carefully chosen parameters, the hybrid forecasting model may not be better than a single forecasting model. So we combine the CPSO-BP neural network with input parameters selection method to achieve better predict performance.2) The data processing of input dataset. We propose a hybrid prediction model by combining wavelet transfonn with the previous study to ensure the predictive ability and the stability of the model. In the spatial-temporal forecasting analysis, this study presents the idea of projection:the core idea is to simplify complex issues, achieved by the N-dimensional space to a plurality of one-dimensional mapping, not only reduces the computational difficulty, but also reduce the computation time, thereby reducing the time complexity of the algorithm. On this basis, this paper proposes a projection-based hotspot analysis model, then discusses the time and space complexity of the algorithm from the theoretical level. This model contains both the RNNH accurate prediction feature and lower time cost of K-means algorithm.The target of this study is to build public crisis management case knowledge platform based on the forecasting model.Based on the above theoretical and applied research background, three models established in this study:1) Considering the uncertainty of the earthquake occurd, the dispersion characteristics of the H1N1outbreak areas and the crime in the place of uncertainty, this research combined projection ideas and hierarchical clustering ideas, and then proposed a projection-based hotspot analysis forcasting model. This model has been successfully applied to Auckland crime analysis, seismic data analysis of the United States, H1N1occurred hotspot analysis and global seismic data analysis.2) The research background was set in energy crisis, back propagation neural network based on particle swam optimization combine with input parameters selcetion model has been applied to the wind speed forecasting, through experimental results, it has been verified the validity of the prediction model.3) Considering the importance of the influenza forecasting, this study used BP neural network based on particle swam optimization combine with wavelet analysis into the United States, Canada, Australia and South Africa influenza forecasting,then described from the prediction accuracy and stability of the algorithm.The main research achievements and contributions are as follows:1) Applied set theory in the field of forcasting, and used it as a basis analysis tools for SOA-based public crisis case management platform.2)In the field of spatial-temporal analysis, proposed projection-based hotspot analysis forecasting model, and successfully applied it into predict infectious disease, crime analysis and seismic hotspots analysis.3) Researched on two new hybrid forecasting model, then successfully applied them into wind speed prediction and influenza forecasting to verify the validity of the models.4) Building a SOA-based public crisis knowledge management platform, make the forecasting models successfully applied to the actual engineering field.