Abstract:

With the development of traffic flow detection technology, massive traffic flow information can be obtained more easily and efficiently. Aiming at the accuracy requirement of short-term traffic flow prediction, an optimization method of short-term traffic flow prediction based on deep learning is proposed. The Long Short-Term Memory algorithm of neural network is used to process the data with the idea of multi-factor analysis. Through the multi-factor analysis of short-term traffic flow data, such as weather factors, holidays, etc., short-term traffic flow data are divided into multiple data sets according to factors, and different data sets are used as training sets to preview. The traffic flow in the next few days with the same factors as the training set is measured. Through this method, the obtained data are more pure, and the influence of various factors on traffic flow prediction is effectively solved. The results show that the optimization method can better reflect the characteristics of road traffic flow change and overcome the disadvantage that the factors of traffic flow data set are not single.

Key words: intelligent transportation system, short-term traffic flow, multivariate analysis, Long Short-Term Memory

摘要：

随着交通流检测技术的发展，海量的交通流信息可以更容易高效地获取，针对短时车流量预测的准确性要求，提出了一种结合深度学习的短时车流量预测优化方法，采用神经网络Long Short-Term Memory算法，用多因素分析的思想对数据进行处理。通过对短时交通流数据进行多因素分析，如天气因素、节假日等，将短时交通流数据划分为多种数据集，将划分的不同数据集作为训练集去预测与训练集因素相同的未来时刻车流量情况。通过这种方法，使得获取的数据更为纯净，有效解决了多种因素对车流量预测影响问题。结果表明，该优化方法克服了车流量数据集影响因素不单一的缺点，能够更为准确地反映道路交通流的变化特征。

关键词: 智能交通系统, 短时车流量, 多因素分析, Long Short-Term Memory