基于粗集和神经网络耦合的短时交通流预测

比较分析神经网络和粗糙集在数据处理过程中的各自优缺点,提出一种基于二者强耦合集成方式的短时交通流预测模型。首先利用粗集对获取的交通流数据进行预处理,简化神经网络训练样本数据集并通过粗集属性约简提取决策规则;其次,利用所提取的规则直接确定神经网络的隐层数、隐层节点数及节点的相互关系;最后训练神经网络用于短时交通流预测。通过与单纯利用神经网络预测的结果进行比较,发现该模型降低了网络训练时间,提高了预测精度。

Short-term Traffic Flow Forecasting Based on Coupling of Rough Set and Neural Network

While comparing the features of data processing by neural network and rough set, a model based on strong coupling of rough set and neural network was proposed for forecasting short-term traffic flow. First, the traffic flow data were pre-processed by rough set to simplify the neural network training data set and extract the decision rules by rough set attribute reduction. Second, the numbers of hidden layers, hidden layer notes and relationship between the hidden layer notes were fixed according to the extracted rules. Third, the neural network was trained for short-term traffic flow forecasting. By comparing the results concluded by neural network and integrated model, it is confirmed that the integrated model could save training time and improve forecasting accuracy.