基于时间序列相似性搜索的交通流短时预测方法

为了进一步提高交通流短时预测的效果,在分析现有预测模型存在问题的基础上,设计了1种基于时间序列相似性搜索的交通流短时多步预测方法.利用界标模型对交通流时间序列数据进行模式表示,在历史数据库中搜索与当前交通流时间序列相似度较高的历史时间序列,进而确定与预测时刻相对应的历史数据,利用回声状态网络模型实现交通流的短时多步预测.采用某特大城市快速路5 min采样间隔的交通流量数据进行实验验证和对比分析.实验结果表明,回声状态网络模型的预测精度分别比ARIMA模型和BP神经网络模型提高了6.25%和3.85%,以时间序列相似性搜索结果作为模型输入数据能够进一步提高交通流短时预测的精度. 

A Short-term Traffic Flow Prediction Method Based on Similarity Search of Time Series

In order to improve the effectiveness of short-term traffic flow prediction, in view of the existing forecasting models, this paper presents a short-term traffic flow multi-step prediction method based on similarity search of time series. Firstly, the landmark model is used to represent time series of traffic flow data. Then, the input data of prediction model are determined through searching similar time series. Finally, an echo state networks model is used for traffic flow multi-step prediction. The performance of the proposed method is evaluated with expressway traffic flow data collected from loop detectors in a Chinese megacity. The experimental results demonstrate that the prediction accuracy of the echo state networks model is 6.25% and 3. 85% higher than ARRIMA model and BP neural network model, respectively. Therefore, using time series similarity search results as input data, the model can further improve the short-term traffic flow prediction accuracy.