Scholarlink

This work contributes to the significant potential of Long Short-Term Memory (LSTM) for traffic flow prediction, which is highly dependent on complex; nonlinear and stochastic data. We propose an LSTM hyperparameters optimization approach using a variant of the Adaptive Reinitialized Differential Evolution (ARDE) algorithm; handling mixed types of genes and boundary constraints. Experiments of the proposed mixed-ARDE-based LSTM algorithm on three analyzed datasets demonstrate a considerable accuracy compared to baseline models, namely DE-LSTM, GA-LSTM and keras-tuner-LSTM.

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

Constrained Mixed-ARDE for Optimizing LSTM Hyperparameters in Traffic Flow Forecasting

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Meriem Talai,
  • Nora Taleb,
  • Hafed Zarzour,
  • Iyad Kouloughli,
  • Mohand Kouloughli

摘要

This work contributes to the significant potential of Long Short-Term Memory (LSTM) for traffic flow prediction, which is highly dependent on complex; nonlinear and stochastic data. We propose an LSTM hyperparameters optimization approach using a variant of the Adaptive Reinitialized Differential Evolution (ARDE) algorithm; handling mixed types of genes and boundary constraints. Experiments of the proposed mixed-ARDE-based LSTM algorithm on three analyzed datasets demonstrate a considerable accuracy compared to baseline models, namely DE-LSTM, GA-LSTM and keras-tuner-LSTM.