基于因果卷积和 Informer 模型的城市公交客流预测

刘妙男, 王魏, 胡显辉, 许德昊

控制工程 ›› 2024, Vol. 31 ›› Issue (8) : 1445-1454.

控制工程 ›› 2024, Vol. 31 ›› Issue (8) : 1445-1454.

基于因果卷积和 Informer 模型的城市公交客流预测

作者信息 +

Urban Bus Passenger Flow Forecasting Based on Causal Convolution and Informer Model

Author information +
文章历史 +

摘要

为了有效缓解日常交通拥堵,倡导健康绿色出行,通过城市客流特征分析和预测精准掌握城市交通中的客流量分布特点,综合考虑天气、节假日等多种影响因素对客流量的影响,基于卷积神经网络提出了一种因果卷积自注意力的城市公交客流量 CCN-Informer 预测模型,改善输入序列局部信息提取能力,获取长期依赖信息,从而提高预测准确性。实验使用因果卷积自注意力进行依赖性特征提取,通过控制编码器的样本计算成本,采用生成式解码器提高模型预测效率。结果表明,CCN-Informer 模型考虑周围重要特征因素影响,样本数据拟合程度明显提高,与基线模型对比,标准误差下降范围为 4%~20%,有更高的预测精度和运行效率,验证了模型的有效性,可以为城市交通客流量预测提供依据。

Abstract

In order to effectively alleviate daily traffic congestion, advocate healthy and green travel, and accurately grasp the distribution characteristics of passenger flow in urban traffic through the analysis and prediction of urban passenger flow characteristics, this paper comprehensively considers the influence of weather, holidays and other factors on passenger flow. The product neural network proposes a causal convolutional self-attention urban bus passenger flow, which improves the local information extraction ability of the input sequence and obtains long-term dependency information, thereby improving the prediction accuracy. In the experiment, causal convolutional self-attention is used for dependency feature extraction, the sample calculation cost of the encoder is controlled by the CCN-Informer model, and the generative structure decoder is used to improve the model prediction efficiency. The results show that the CCN-Informer model considers the influence of the surrounding important characteristic factors, and the fitting degree of the sample data is significantly improved. Compared with the baseline model, the standard error decreases between 4% and 20%, and it has higher prediction accuracy and operation efficiency. The validity of the model is verified, and it can provide a basis for the prediction of urban traffic passenger flow.

关键词

Informer / 因果卷积 / 自注意力机制 / 长时间序列 / 客流量预测

Key words

Informer / causal convolution / self-attention mechanisms / long time series / passenger flow forecast

引用本文

导出引用
刘妙男, 王魏, 胡显辉, 许德昊. 基于因果卷积和 Informer 模型的城市公交客流预测[J]. 控制工程, 2024, 31(8): 1445-1454
LIU Miaonan, WANG Wei, HU Xianhui, XU Dehao. Urban Bus Passenger Flow Forecasting Based on Causal Convolution and Informer Model[J]. Control Engineering of China, 2024, 31(8): 1445-1454

2

Accesses

0

Citation

Detail

段落导航
相关文章

/