【摘要】 随着我国经济的发展，城市轨道交通已经成为公共交通的重要部分。对城市轨道交通的短时客流进行预测，有助于提高轨道交通的服务质量。由于地铁客流分布具有随机性，传统模型往往难以捕获其时间特征。因此提出了一种组合模型，将时序卷积网络（TCN）与ITransformer模型进行串联，以自注意力机制为核心，构建了TCN-ITransformer模型。为验证模型有效性，以上海地铁9号线的进站客流为研究对象，使用K-means聚类算法对21个站点进行聚类，并对聚类后的站点分别进行客流预测。选取LSTM、Informer和ITransformer作为基线模型进行对比验证。实验结果表明，对于时间周期性较强的进站客流数据，模型的均方误差和平均绝对误差优于基线模型。更多还原

【Abstract】 With the development of China’s economy,urban rail transit has become an important part of public transportation.Predicting the short-term passenger flow of urban rail transit can help to improve the service quality of rail transit.Due to the randomness of subway passenger flow distribution,traditional models often struggle to capture its temporal characteristics.This article proposes a combined model that concatenates the Time Series Convolutional Network(TCN)with the ITransformer model,with Self-Attention Mechanism as the core,to construct the TCN-ITransformer model.To verify the effectiveness of the model in this article,the inbound passenger flow of Shanghai Metro Line 9 was taken as the research object.The K-means clustering algorithm was used to cluster 21 stations,and passenger flow predictions were made for each station after clustering.Selecting LSTM,Informer,and ITransformer as baseline models for comparative verification.The experimental results show that for inbound passenger flow data with strong time periodicity,the mean square error and mean absolute error of our model are better than those of the baseline model.更多还原