基于站点实时关联度的短时公交客流预测方法

为探究公交站点之间的关联度并对公交客流进行更精准的实时预测，本文提出基于 Attention的交通预测核心算法(Traffic Forecast Model Based Attention，TFMA)，结合数据预处理和 站点信息编码完成基于站点实时关联度的短时公交客流预测方法。该方法首先创新性地提出了 站点实时关联度，可实现对目标站点客流量更精准的预测；其次，在公交站点的编码信息中融入 线路站点信息、客流变化率、天气、日期等关联因素；接着，该方法依靠Attention机制计算站点实 时关联度；核心算法中使用multi-headed机制、增加通道和残差连接进一步提升预测能力；最后， 以苏州市公交数据进行验证。结果显示：在准确率上，对比多元线性回归的53.8%、GRU(Gated Recurrent Unit)的66.9%和LightGBM(Light Gradient Boosting Machine)的81.2%，本文提出的基于 站点实时关联度的短时公交客流预测方法的准确率在90%以上，表明该方法具备优秀的短时公 交客流预测能力。

Short-term Public Traffic Passenger Volume Forecasting Method Based on Real-time Relevance of Stations

In order to explore the relevance between bus stations and make real-time predictions of bus passenger volume more accurate, this paper proposes a core traffic prediction algorithm based on Attention, referred to as TFMA, which combines data preprocessing and station information coding. A short-term public traffic passenger volume forecasting method based on the real-time relevance of stations is proposed. This method firstly proposes the real-time relevance of stations in an innovative way, which can achieve a more accurate prediction of the passenger volume of the target station. Secondly, this paper integrates the relevant factors(e.g., line station information, passenger volume rate of change, weather) and date into the coding information of the bus station. Then, the method relies on the Attention mechanism to calculate the real-time relevance of the station; the core algorithm also uses the multi-headed mechanism, adding channels and residual connections to further improve the prediction ability. Finally, this paper uses the data of Suzhou city bus for verification. The results show that: in terms of accuracy, compared with 53.8% of multiple linear regression, 66.9% of GRU and 81.2% of LightGBM, the accuracy of the forecasting method proposed in this paper is above 90%, indicating that the method has excellent short-term bus passenger flow prediction capabilities.