基于量子神经网络的道路交通事故预测

道路交通事故预测是道路交通安全研究的一项重要内容. 针对BP神经网络在道路交通事故预测中精度不足及收敛速度慢的问题，引入量子神经网络并构建道路交通事故预测模型. 模型通过对道路交通事故时间序列进行相空间重构，有效扩充训练样本数量；且隐含层神经元采用态叠加的激励函数，对道路交通事故数据的特征空间进行多层梯级划分，以快速匹配输入数据与特征空间的对应关系，提高模型的收敛速度；在训练过程中动态调整量子间隔，以响应事故数据的强随机性. 实验结果表明，该预测模型能够较好地适应道路交通事故数据的特性，且预测精度和收敛速度较改进BP神经网络有显著提高.

Road Traffic Accidents Forecasting Based on Quantum Neural Network

Road traffic accidents forecasting is one of the key issues of the road traffic safety. A forecasting model based on the quantum neural network (QNN) is introduced to solve the problems of precision inadequacy and low convergence rate of the BP neural network. The numbers of the training samples are enlarged by phase-space reconstruction of the road traffic accidents time series. The activation function of model superposition is used in the hidden-layer neurons to partition the feature space of road traffic accidents data into multi-layer rundles. Both the matching speed of the input data and feature space and the convergence rate of model are improved. The quantum interval is adjusted dynamically in the training process to adapt the strong randomness. The results show that the QNN model fits the characteristics of road traffic accidents data well, and its prediction accuracy and convergence rate are superior to that of improved BP neural network.