Scholarlink

Efficient traffic forecasting is imperative for optimizing supply chain management (SCM) by identifying optimal transport routes and minimizing logistical costs. This study proposes a fine-tuned Long Short-Term Memory (LSTM) model designed to outperform Temporal Convolutional Networks (TCN) in traffic volume prediction tasks. Leveraging the Metro Interstate Traffic Volume Dataset, the LSTM model demonstrates a 15% improvement in Mean Squared Error (MSE) over TCN, effectively capturing long-term temporal dependencies. The model achieves superior forecasting accuracy by employing advanced preprocessing techniques, such as normalization and categorical encoding, and optimizing hyperparameters to mitigate overfitting. Experimental results validate the model’s capability to provide actionable insights for transportation planning, leading to reduced fuel consumption and enhanced delivery efficiency. This research fills a critical gap by addressing the limitations of existing models in real-time SCM applications, making the LSTM model a viable tool for real-world traffic prediction and logistical optimization. Future research could explore hybrid models and incorporate real-time data to further improve prediction robustness.

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

A Case Study on Traffic Forecasting for Supply Chain Management: Optimization of the Long Short-Term Memory Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Vasanthi Govindaraj,
  • Prakash Periyasamy

摘要

Efficient traffic forecasting is imperative for optimizing supply chain management (SCM) by identifying optimal transport routes and minimizing logistical costs. This study proposes a fine-tuned Long Short-Term Memory (LSTM) model designed to outperform Temporal Convolutional Networks (TCN) in traffic volume prediction tasks. Leveraging the Metro Interstate Traffic Volume Dataset, the LSTM model demonstrates a 15% improvement in Mean Squared Error (MSE) over TCN, effectively capturing long-term temporal dependencies. The model achieves superior forecasting accuracy by employing advanced preprocessing techniques, such as normalization and categorical encoding, and optimizing hyperparameters to mitigate overfitting. Experimental results validate the model’s capability to provide actionable insights for transportation planning, leading to reduced fuel consumption and enhanced delivery efficiency. This research fills a critical gap by addressing the limitations of existing models in real-time SCM applications, making the LSTM model a viable tool for real-world traffic prediction and logistical optimization. Future research could explore hybrid models and incorporate real-time data to further improve prediction robustness.