遗传算法和神经网络在交通事故预测中的应用

介绍了BP(误差反向传播算法)和GA(遗传算法)以及GA-BP三种神经网络,并以此分别对道路交通事故进行预测。实验结果表明,基于GA-BP算法的神经网络方法应用于交通事故的预测问题,能采用遗传学习算法优化BP神经网络模型的初始权重,即先利用遗传学习算法进行全局训练,再用BP算法进行精确训练,使网络收敛速度加快和避免局部极小。GA-BP神经网络在收敛速度和预测精度等方面均优于BP和GA网络,从而为交通部门的未来事故预测提供了一种新的思路与方法。

Application of genetic algorithm and neural network in the traffic accident forecast

Three kinds of neural networks of BP(back programmes)and GA(genetic algorithm)as well as GA-BP are introduced,and the road traffic accident is separately forecast.The experimental result show that,the nural network method based on the GA-BP algorithm is applied to forecast the traffic accidents,and can optimize the initial weight of BP neural network model using the genetic study algorithm.Namely first carries on the overall situation training using the genetic study algorithm,then uses the BP algorithm to carry on the precise training,causes the network convergence rate raised and partial to avoid being minimum.GA-BP neural network surpasses BP and the GA network in aspects of convergence rate and forecast precision,thus it has provided one kind of new mentality and the method for future accident forecast of traffic department.