Scholarlink

<p>Accurate short-term traffic flow prediction is crucial for the development of intelligent transportation systems (ITS), yet it remains challenging due to the stochasticity, spatiotemporal complexity, and sensitivity to external disruptions of traffic patterns. To address these issues, this paper proposes IVY-LSTM, a novel hybrid model that integrates a biologically inspired Ivy Growth Optimization Algorithm (IVYA) with Long Short-Term Memory (LSTM) networks. IVYA adaptively optimizes key LSTM hyperparameters–including the number of neurons, dropout rate, batch size, and learning rate–by simulating ecological behaviors of ivy plants such as phototropism, clonal propagation, and natural selection. This approach effectively balances exploration and exploitation in high-dimensional search spaces. We evaluate IVY-LSTM on two real-world datasets, including hourly traffic flow data from New York (10,000 samples), and demonstrate superior performance over LSTM, xLSTM, sLSTM, mLSTM, and PSO-LSTM. IVY-LSTM achieves a 19.6% reduction in MAE (4.05 vs. 5.039), a 19.5% reduction in RMSE (5.35vs.6.65), and an improved <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^2\)</EquationSource> </InlineEquation> score of 0.901, underscoring its robustness and potential for real-time traffic management in smart cities.</p>

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 bio-inspired IVY metaheuristic for optimizing LSTM models in urban traffic flow prediction

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Bo Liang,
  • Jiaru Deng,
  • Fujiang Yuan,
  • Jie Gao,
  • Junye Yang,
  • Peng Zhao

摘要

Accurate short-term traffic flow prediction is crucial for the development of intelligent transportation systems (ITS), yet it remains challenging due to the stochasticity, spatiotemporal complexity, and sensitivity to external disruptions of traffic patterns. To address these issues, this paper proposes IVY-LSTM, a novel hybrid model that integrates a biologically inspired Ivy Growth Optimization Algorithm (IVYA) with Long Short-Term Memory (LSTM) networks. IVYA adaptively optimizes key LSTM hyperparameters–including the number of neurons, dropout rate, batch size, and learning rate–by simulating ecological behaviors of ivy plants such as phototropism, clonal propagation, and natural selection. This approach effectively balances exploration and exploitation in high-dimensional search spaces. We evaluate IVY-LSTM on two real-world datasets, including hourly traffic flow data from New York (10,000 samples), and demonstrate superior performance over LSTM, xLSTM, sLSTM, mLSTM, and PSO-LSTM. IVY-LSTM achieves a 19.6% reduction in MAE (4.05 vs. 5.039), a 19.5% reduction in RMSE (5.35vs.6.65), and an improved \(R^2\) score of 0.901, underscoring its robustness and potential for real-time traffic management in smart cities.