Application Research Of Particle Swarm Optimization BP Neural Network On Emergency Resource Demand Forecasting

In recent years, frequent occurrence of various natural disasters and emergencies causedgreat threaten on life and property of human society. Emergency supplies is the basic bulwarkof rescue work, scientific reasonable emergency supplies demand forecasting results is thekey to high efficient rescue, which can provide the basic data for the organization andscheduling of emergency supplies. In view of this, this thesis conducts the following research:Firstly, this thesis analyses and summarizes the contents and characteristics of theemergency supplies demand based on concepts of emergency supplies demand. Uses FuzzyAnalytic Hierarchy Process to calculate the critical factors’ fuzzy weight, expert assign valuesof key factors based on field information which gathered by rescue command subsystem inreal time, then form factor fuzzy matrix, calculate the Hamming distance to get the event level,and the event level can provide the reference for the supplies demand forecasting.Secondly, this thesis introduces BP neural network and PSO algorithm. As BP neuralnetwork is easy to fall into local minimum and slow convergence speed and other defects,thesis uses PSO to optimize BP neural network, which has the characteristics of globaloptimization ability and fast convergence speed, the weight and threshold of optimizationneural network can converge to the optimum solutions quickly. Then the strong local searchcapability of BP neural network can be further used to find optimal solution, so the optimalvalue of weights and thresholds of BP network can be obtained. The fusion algorithm isapplied in emergency supplies demand forecasting.Finally, an emergency supplies demand forecast model is established based on PSO-BPneural network, then uses the actual situation of emergency supplies demand as the researchbackground, conducts a simulation research and comparative analysis. The result shows thatcompared with individual BP network, PSO-BP network prediction method is much easier toavoid local optimum solution and convergence much faster. It improves rate of prediction andoffers an effective approach for actual emergency rescue.