Scholarlink

<p>The precise prediction of greenhouse temperatures is crucial for the optimization of crop cultivation and resource allocation. Yet, the dynamic nature of greenhouse climates challenges conventional predictive models. This research overcomes the difficulties by creating hybrid machine learning (ML) models—namely, Radial Basis Function (RBF), Random Forest (RF), and Extreme Gradient Boosting (XGBoost)—that have been optimized with the Slime Mold Algorithm (SMA) and Moth-Flame Optimization (MFO) for improved predictive accuracy. The results show that models optimized by MFO methods outperform models optimized by SMA, with XGBoost-MFO achieving the best accuracy (<i>R</i><sup>2</sup> = 0.974861, MSE = 0.797552). These results highlight the advantages of using intelligent optimization strategies to improve model flexibility and performance. In addition to predictive improvements, this study examines the practical implementation of these models in real-world greenhouse operations. Artificial intelligence-controlled temperature can improve energy efficiency, automate climate control, and enhance resource utilization. However, issues such as sensor reliability, data acquisition costs, and computational demands must be addressed to make application mainstream. This study highlights the significance of hybrid machine-learning models in sustainable agriculture by connecting predictive analytics with real-world applications. The results presented herein facilitate the development of automated decision-making systems for the control of greenhouses that are climate change resilient.</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

Enhancing the prediction accuracy of greenhouse temperatures through hybrid machine learning models: an in-depth comparative analysis of RBF, RF, and XGBoost techniques combined with SMA and MFO optimization strategies

  • 期刊论文

  • 发表日期：

    2025年5月23日

查看全文
  • Zhenhua Dai,
  • Tangsen Huang

摘要

The precise prediction of greenhouse temperatures is crucial for the optimization of crop cultivation and resource allocation. Yet, the dynamic nature of greenhouse climates challenges conventional predictive models. This research overcomes the difficulties by creating hybrid machine learning (ML) models—namely, Radial Basis Function (RBF), Random Forest (RF), and Extreme Gradient Boosting (XGBoost)—that have been optimized with the Slime Mold Algorithm (SMA) and Moth-Flame Optimization (MFO) for improved predictive accuracy. The results show that models optimized by MFO methods outperform models optimized by SMA, with XGBoost-MFO achieving the best accuracy (R2 = 0.974861, MSE = 0.797552). These results highlight the advantages of using intelligent optimization strategies to improve model flexibility and performance. In addition to predictive improvements, this study examines the practical implementation of these models in real-world greenhouse operations. Artificial intelligence-controlled temperature can improve energy efficiency, automate climate control, and enhance resource utilization. However, issues such as sensor reliability, data acquisition costs, and computational demands must be addressed to make application mainstream. This study highlights the significance of hybrid machine-learning models in sustainable agriculture by connecting predictive analytics with real-world applications. The results presented herein facilitate the development of automated decision-making systems for the control of greenhouses that are climate change resilient.