Abstract: Traffic congestion prediction is an important part of smart traffic, but a large amount of traffic data cannot be obtained in a public way. Based on incomplete data, a method based on spatial-temporal correlativity for traffic congestion prediction is proposed. The improved kernel density estimation method is adopted, so that the prediction process does not rely on a large amount of historical data for training, and the traffic congestion can be accurately predicted in real time with the partially collected data. The proposed traffic congestion prediction method is validated in the real data set, and the experimental results show the feasibility of this method in real-time traffic prediction.

Key words: traffic congestion, spatial-temporal correlativity, kernel density estimation, real-time prediction, incomplete data

摘要： 交通拥堵预测是智慧交通一个重要组成部分，但是大量的交通数据无法以公开的方式获取。在不完全数据下，提出了一种基于时空相关性的交通拥堵预测方法。该方法采用改进的核密度估计法，使得预测过程中不依赖大量历史数据进行训练，直接利用部分采集到的数据精准地实时地对交通拥堵进行预测。在真实数据集上对提出的交通拥堵预测方法进行验证，实验结果表明了该方法在实时交通预测上的可行性。

关键词: 交通拥堵, 时空相关性, 核密度估计, 实时预测, 不完全数据