Scholarlink

The prediction of shared bicycle demand is a cornerstone for enterprises in the sector to operate scientifically, manage effectively, and allocate regional resources dynamically. This process plays a pivotal role in optimizing the supply of services. To predict shared bicycle demand with precision, this study utilizes shared bicycle data from Shenzhen, rasterizes the study area, and examines the significant factors influencing users’ travel choices from various perspectives, including road network conditions, public transportation infrastructure, points of interest, population size, and other built-environment aspects. This study introduces a CNN-LSTM-Attention model tailored for shared bicycle demand prediction analysis. Our findings indicate that the CNN-LSTM-Attention model exhibits a coefficient of determination (R2) of 0.97214, which surpasses the prediction accuracy of the CNN, LSTM, and CNN-LSTM models. This high level of accuracy satisfies the demands of precise prediction and further validates the rationality of our proposed prediction model. Consequently, this work offers theoretical guidance for shared bicycle enterprises to achieve optimal resource allocation.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Study on Urban Shared Bike Demand Forecasting Based on the CNN-LSTM-Attention Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Siqi Zhang,
  • Xiaohan Kou,
  • Jianying Zhou,
  • Xiaoyu Wang

摘要

The prediction of shared bicycle demand is a cornerstone for enterprises in the sector to operate scientifically, manage effectively, and allocate regional resources dynamically. This process plays a pivotal role in optimizing the supply of services. To predict shared bicycle demand with precision, this study utilizes shared bicycle data from Shenzhen, rasterizes the study area, and examines the significant factors influencing users’ travel choices from various perspectives, including road network conditions, public transportation infrastructure, points of interest, population size, and other built-environment aspects. This study introduces a CNN-LSTM-Attention model tailored for shared bicycle demand prediction analysis. Our findings indicate that the CNN-LSTM-Attention model exhibits a coefficient of determination (R2) of 0.97214, which surpasses the prediction accuracy of the CNN, LSTM, and CNN-LSTM models. This high level of accuracy satisfies the demands of precise prediction and further validates the rationality of our proposed prediction model. Consequently, this work offers theoretical guidance for shared bicycle enterprises to achieve optimal resource allocation.