Scholarlink

<p>This study proposes an enhanced Dwarf Mongoose Optimization Algorithm(DMOA) that integrates an Intuitionistic Fuzzy Entropy Perturbation Convergence Factor and a Hunger-driven Search Strategy. By incorporating the population’s Intuitionistic Fuzzy Entropy (IFE) as a perturbation factor into the convergence mechanism, the algorithm adaptively balances local search and global exploration based on the population’s degree of aggregation. Additionally, a hunger-based regulation strategy is introduced, allowing the algorithm to dynamically adjust the search space, particularly during later iterations, thereby enhancing individual local search capabilities and effectively avoiding local optima. To improve the uniform distribution of the population in the solution space, the Henon chaotic mapping is employed for population initialization. Experimental results on multiple CEC benchmark functions demonstrate that the proposed IFSDMOA algorithm excels in convergence speed, solution accuracy, and robustness. Finally, the IFSDMOA algorithm is applied to the hyperparameter optimization of a Transformer neural network to enhance the performance of a stock market prediction model. Comparative results with other time series prediction models indicate that IFSDMOA-Transformer exhibits superior predictive accuracy and generalization capability in forecasting stock market prices.</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

Time series prediction with transformer neural network optimized by IFE and hunger-driven DMOA

  • 期刊论文

  • 发表日期：

    2025年8月30日

查看全文
  • Shixing Zheng,
  • Suyang Ma,
  • Xu Jin

摘要

This study proposes an enhanced Dwarf Mongoose Optimization Algorithm(DMOA) that integrates an Intuitionistic Fuzzy Entropy Perturbation Convergence Factor and a Hunger-driven Search Strategy. By incorporating the population’s Intuitionistic Fuzzy Entropy (IFE) as a perturbation factor into the convergence mechanism, the algorithm adaptively balances local search and global exploration based on the population’s degree of aggregation. Additionally, a hunger-based regulation strategy is introduced, allowing the algorithm to dynamically adjust the search space, particularly during later iterations, thereby enhancing individual local search capabilities and effectively avoiding local optima. To improve the uniform distribution of the population in the solution space, the Henon chaotic mapping is employed for population initialization. Experimental results on multiple CEC benchmark functions demonstrate that the proposed IFSDMOA algorithm excels in convergence speed, solution accuracy, and robustness. Finally, the IFSDMOA algorithm is applied to the hyperparameter optimization of a Transformer neural network to enhance the performance of a stock market prediction model. Comparative results with other time series prediction models indicate that IFSDMOA-Transformer exhibits superior predictive accuracy and generalization capability in forecasting stock market prices.