针对城市大规模路网交通状态在线预测问题,采集出租车上的全球定位系统数据作为浮动汽车数据(FCD),采用广义神经网络理论,应用网格并行算法,通过加速广义神经网络的收敛和学习速度,构建了在线交通状态预测新模型.实验结果表明,该算法实现了在线广义神经网络交通状态预测模型的并行计算,提高了预测模型的计算速度,满足大规模交通路网实时、高效的在线交通状态预测的实际需求. In order to solve the problem of large-scale on-line forecasting of traffic conditions in urban road network, GPS data of taxi are collected as floating car data (FCD). By using generalized neural network theory and grid parallel algorithm, by accelerating generalized neural network convergence and learning Speed ​​and build a new model of online traffic state prediction.The experimental results show that this algorithm realizes the parallel computing of traffic state prediction model of online generalized neural network and improves the computational speed of the predictive model to meet the requirements of large-scale real-time and efficient online traffic network Actual demand for traffic status prediction.