摘要： 针对高速公路交通事故引发交通堵塞的问题，提出一种基于减法聚类和自适应神经模糊推理系统的事件持续时间预测新方法。将该方法应用于交通事件持续时间预测，从I-880数据库中提取事件持续时间相关因素，使用非参数估计法进行显著性分析，将影响程度最大的因素作为模糊系统的输入样本，采用减法聚类对输入样本进行聚类，得到模糊规则数并建立初始模糊推理系统，使用BP反向传播算法和最小二乘估计算法的混合算法对该模糊系统进行训练并优化，建立最终模糊模型。仿真结果证明，该系统对交通事件持续时间预测具有较高检测率和较低误报率。
关键词:
事件持续时间,
方差分析,
减法聚类,
自适应神经模糊推理系统,
ROC曲线
Abstract: Aiming at the problems of traffic jams caused by highway traffic accident and improving highway operation safety, this paper proposes a subtraction clustering method combined with Adaptive Neural-fuzzy Inference System(ANFIS) applied in traffic incident duration prediction. Forecasting process is that extracting duration event related factors from I-880 database, using a parameter estimation method for significant analysis, choosing bigger factors as fuzzy system’s input samples, the subtractive clustering is introduced to confirm the fuzzy rule number to build the initial fuzzy inference system, the hybrid algorithm is used to train and optimize the fuzzy system and establish a final training fuzzy model. Simulation results show that the system for traffic incident duration prediction has higher detection rate and lower false positives, in general.
Key words:
incident duration,
analysis of variance,
subtractive clustering,
Adaptive Neural-fuzzy Inference System(ANFIS),
ROC curve
中图分类号: