Scholarlink

<p>Accurate prediction of industrial added value growth rate is critical for assessing regional economic performance, especially amid geopolitical uncertainty and public health crises. This paper proposes a Hybrid Multi-Scale Optimization Forecasting Model for the industrial added value growth rate that integrates multiple advanced techniques. The model first employs the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) to decompose the time series into intrinsic mode functions representing different frequency components. Subsequently, the Binary Harris Hawks Optimizer (BHHO) performs feature selection from the candidate variables, while the Grey Wolf Optimizer (GWO) is used to fine-tune the hyperparameters of the Support Vector Regression (SVR) model. The optimized model is then used for prediction. Empirical results demonstrate that the proposed model effectively captures the fluctuation trends of the industrial added value growth rate and significantly outperforms other traditional models in terms of predictive accuracy, achieving RMSE values of 1.28, 1.94, and 5.37 and MAE values of 0.76, 1.54, and 4.64 on the training, validation, and test sets, respectively. The findings provide scientific evidence and reference for regional economic policy formulation and industrial planning in Xinjiang.</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

The Hybrid Multi-Scale Optimization Forecasting Model: a Novel Framework for High-Accuracy Prediction of Industrial Added Value Growth Rate

  • 期刊论文

  • 发表日期：

    2025年12月23日

查看全文
  • Xiaoxian Duan,
  • Kun Guo

摘要

Accurate prediction of industrial added value growth rate is critical for assessing regional economic performance, especially amid geopolitical uncertainty and public health crises. This paper proposes a Hybrid Multi-Scale Optimization Forecasting Model for the industrial added value growth rate that integrates multiple advanced techniques. The model first employs the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) to decompose the time series into intrinsic mode functions representing different frequency components. Subsequently, the Binary Harris Hawks Optimizer (BHHO) performs feature selection from the candidate variables, while the Grey Wolf Optimizer (GWO) is used to fine-tune the hyperparameters of the Support Vector Regression (SVR) model. The optimized model is then used for prediction. Empirical results demonstrate that the proposed model effectively captures the fluctuation trends of the industrial added value growth rate and significantly outperforms other traditional models in terms of predictive accuracy, achieving RMSE values of 1.28, 1.94, and 5.37 and MAE values of 0.76, 1.54, and 4.64 on the training, validation, and test sets, respectively. The findings provide scientific evidence and reference for regional economic policy formulation and industrial planning in Xinjiang.