Research On Short-term Traffic Flow Prediction Algorithm Based On Optimized BP Neural Network

In the increasingly busy work, transportation is playing a more and more important role in our daily lives. Alleviating city traffic jams, improving efficiency in the use of road network and trying our best to reduce the rate of traffic accident have become the consensus of all countries. The road network traffic flow prediction algorithm which further research is needed has been used to solve a series of problems mentioned above, especially plays an increasingly important part in the intelligent transportation systems(ITS).Based on BP neural network, this paper established a short-term traffic flow forecasting model. Then the genetic algorithm and mind evolution algorithm were used to optimize the BP. Different from genetic algorithm, mind evolution algorithm uses "group optimization" instead of "individual optimization". It uses "convergence" and "alienation" to replace the operation of "crossover" and "mutation" in GA. Firstly, the BP neural network algorithm which is optimized by mind evolution algorithm should produce initial population. Then the weights and thresholds of BP neural network are optimized. When the training is over, the algorithm will evaluate the training results. If it does not get the desired results, the algorithm would optimize the operation again. In this paper, we analyzed the principles of three kinds of models and carried on the simulation with the actual traffic data. The simulation results were compared among three algorithms. Experimental test results showed that the short-time traffic flow forecasting model based on BP which was optimized by mind evolution algorithm could obtain relatively high accurate forecast value.