基于OD识别和客流预测的城市公共交通多源数据分析关键技术研究

Research on Key Technologies of Multi-Source Data Analysis of Urban Public Transport Based on OD Identification and Passenger Flow Prediction

随着城市公共交通体系的逐步完善,通过大数据等新兴技术手段,深入研判和预测公共交通客流情况至关重要。目前公交OD (origin-destination)识别效率及预测效率都比较有限,客流分配不够合理,无法有效支撑公交出行特征分析、公交专项规划管理等,亟需从理论和实践出发,开展进一步的深入研究和探索,因此需要对现有技术进行改良。本文结合大数据及机器学习等,对公共交通数据进行分析,首先结合Transcad最短路径方法对公交OD识别算法进行优化,其次分别采用五种模型构建公交短时客流预测算法并进行对比分析。结果表明本文在优化公交OD识别算法后,算法效率上有较大提升,同时公交短时客流预测中Tensor + ARIMA方法在效率和精度方面表现较好,能够更好地适用于该领域的实践应用。

With the gradual improvement of the urban public transportation system, it is crucial to study and predict the public transportation passenger flow in depth through big data and other emerging technical means. The current public transportation OD (origin-destination) identification efficiency and prediction efficiency are relatively limited, and the passenger flow distribution is not reasonable enough to effectively support the analysis of public transportation travel characteristics and public transportation special planning management, etc. Further in-depth research and exploration are urgently needed from theory and practice, so the existing technology needs to be improved. This article combines big data and machine learning techniques to analyze public transportation data. Firstly, it optimizes the public transit OD identification algorithm by integrating Transcad’s shortest path method. Secondly, it constructs five different models to develop short-term public transportation passenger flow prediction algorithms and compares them. The results show that after optimizing the public transit OD identification algorithm, there is a significant improvement in algorithm efficiency. Additionally, the Tensor + ARIMA method performs well in terms of efficiency and accuracy for short-term public transportation passenger flow forecasting, making it more suitable for practical applications in this field.