“十四五”期间我国工业发展水平稳步提升，各种工业发展政策的制定以及产业布局都离不开对工业增加值进行定量分析。然而目前针对工业增加值的研究存在些许不足，一是预测工业增加值增速时，普遍存在影响因素选择偏主观以及缺乏对工业增加值影响因子进行对比实证分析等问题，二是在工业增加值预测模型上，参数的选择凭借经验主义，主观性强，影响模型的预测精度。

首先，选取陕西省工业方面相关数据作为数据样本，通过灰色关联分析构建出全面的区域工业增加值影响因素指标体系，陕西省与其他省份的灰色关联分析对比体现出影响指标的选取具有区域性差异。针对工业增速影响因素众多，且存在冗余属性的可能，采用基于知识粒度的粗糙集属性约简对影响因素进行降维，在此基础上提出一种基于RS-SVR工业增加值增速预测模型，将此模型与SVR模型、RS-BP模型进行预测误差和精度方面的对比实验，结果表明：引入粗糙集理论消除了冗余属性，降低了输入维度，有效提升了支持向量回归机(SVR)模型的泛化能力和预测性能。同时，与BP模型相比，SVR模型在工业增速预测方面表现出更好的预测性能。

然后，采用改进的PSO算法AIWPSO对SVR模型进行参数寻优，以进一步提高区域工业增加值增速的预测精度。在此基础上提出了一种基于RS-AIWPSO-SVR工业增加值增速预测模型，使用对比实验以评估提出的RS-AIWPSO-SVR模型与RS-PSO-SVR模型、RS-GA-SVR模型以及使用网格搜索优化的RS-SVR模型在预测误差和精度方面的性能，研究结果表明：RS-AIWPSO-SVR在参数优化方面表现出更快的收敛速度，在预测方面具有最佳的泛化能力、最小的预测误差最小以及最高的预测精度。最后，将目标预测模型与ARIMA模型结合对陕西省未来三个月的工业增加值增速进行预测。综合来看，将粗糙集与改进的PSO算法引入到区域工业增加值增速预测的SVR模型当中，有助于降低预测模型的复杂度和计算难度，具有一定的创新性，有效克服了以往研究的不足和局限，给区域工业增加值增速预测问题提供了新的方法和思路。

During the "14th Five-Year Plan" period, China's industrial development level has seen steady improvement. The formulation of various industrial development policies and layouts crucially relies on quantitative analysis of industrial value added. However, there are some shortcomings in the current research on the value added of industry, one is that in the prediction of the growth rate of value added of industry, two common issues arise: subjectivity in selecting influencing factors and a dearth of comparative empirical analysis regarding factors influencing industrial value added. Additionally, in prediction models for industrial value added, parameter selection relies heavily on empiricism and subjectivity, thereby impacting the accuracy of predictions.

Firstly, the relevant industrial data of Shaanxi Province were selected as data samples, and a comprehensive index system of influencing factors of regional industrial added value was constructed through grey correlation analysis, and the comparison of grey correlation analysis between Shaanxi Province and other provinces reflected that the selection of influencing indicators had regional differences. In view of the fact that there are many influencing factors of industrial growth rate and the possibility of redundant attributes, the rough set attribute reduction based on knowledge granularity is used to reduce the dimensionality of the influencing factors, and on this basis, a prediction model based on RS-SVR industrial added value growth rate is proposed, and the prediction error and accuracy of this model are compared with SVR model and RS-BP model, and the results show that the rough set theory is introduced to eliminate the redundant attributes, reduce the input dimension, and effectively improve the Support Vector Regression Machine (SVR) The generalization ability and prediction performance of the model. At the same time, compared with the BP model, the SVR model shows better prediction performance in industrial growth forecasting.

Then, the improved PS0 algorithm AIWPSO is used for parameter optimization of SVR model to further improve the prediction accuracy of regional industrial value added growth rate. On this basis, a prediction model based on RS-AIWPSO-SVR industrial value added growth rate is proposed, and the RS-AIWPSO-SVR model proposed in this paper is compared with RS-PSO-SVR model, RS-GA-SVR model and RS-SVR model using grid search optimization in terms of prediction error and accuracy. AIWPSO-SVR has faster convergence speed in parameter optimization, the best learning generalization ability in prediction, the smallest prediction error and the best prediction accuracy. Finally, the target forecasting model is combined with ARIMA model to forecast the growth rate of industrial value added in Shaanxi Province in the next three months. In summary, the introduction of rough set and improved PSO algorithm into the SVR model for regional industrial value added growth rate prediction reduces the complexity and computational difficulty of the prediction model, which is innovative, effectively overcomes the shortcomings and limitations of the previous research, and provides new methods and ideas for the regional industrial value added growth rate prediction problem.