Study On Road Traffic Accident Forecasting Based On Quantum Neural Network

Road traffic accident is an important issue in modern transportation system. It not only poses grate threat for people's lives and properties, but also has significant influence on social and ecological environment. The forecasting of road traffic accident is an important research content of road traffic safety, it aims to master the trend of road traffic accident, provide the basis for road traffic management, evaluate rationality of road traffic safety measures and reduce road traffic accident.Based on the status quo of road traffic accident forecasting at home and abroad, considering the characteristics of nonlinear, dynamic and uncertainty of road traffic system, this paper discusses the problems of current forecasting methods, and points out the reason of the problems. Due to lack of full consideration for road traffic accident data, so the convergence rate of forecasting model is slower, and the precision of the model result is lower.Considering the above characteristics of road traffic system, using the advantages of both the artificial neural networks to solve complex nonlinear problems and the neural network of multi-layer activation function to forecast uncertainty and unsteady-state data, this paper presents a new road traffic forecasting model—quantum neural network (QNN) model and studies the key technique of model construction.During designing parameters, empirical questions are considered and a qualitative and quantitative method is presented for selecting parameters of input layer. The method uses element analysis to screen influence factor, and uses correlation analysis to calculate correlation coefficient of variables and result variables. During training network, the local teeny problem is considered and the traditional data preprocessing method is improved. Using this method, the primary samples are normalized to between 0.1 and 0.9, and the local teeny problem is solved.Finally, based on China's population, road mileage, civilian vehicle quantities, GDP and road traffic accident data from 1973 to 2009, this simulation validates the QNN forecasting model and compares with gray model, regression model and BP neural network model. The results show that the forecasting accuracy and convergence rate of QNN model are superior to other models, its feasibility and effectiveness are validated, and it provides reference value for road traffic safety management.