[1]郭茜,赵否曦,彭宇翔,等.基于SCAD先验知识算法的脑卒中预测模型研究[J].计算机技术与发展,2025,(05):136-144.[doi:10.20165/j.cnki.ISSN1673-629X.2025.0005]
 GUO Xi,ZHAO Fou-xi,PENG Yu-xiang,et al.Research on Stroke Incidence Prediction Model Based on SCAD Prior Knowledge Algorithm[J].,2025,(05):136-144.[doi:10.20165/j.cnki.ISSN1673-629X.2025.0005]
点击复制

基于SCAD先验知识算法的脑卒中预测模型研究()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2025年05期
页码:
136-144
栏目:
人工智能
出版日期:
2025-05-10

文章信息/Info

Title:
Research on Stroke Incidence Prediction Model Based on SCAD Prior Knowledge Algorithm
文章编号:
1673-629X(2025)05-0136-09
作者:
郭茜1赵否曦2彭宇翔3支亚京1汪华1王娟1刘国强4*
1. 贵州省气象数据中心,贵州 贵阳 550002;
2. 贵州省疾病与预防控制中心,贵州 贵阳 550002;
3. 贵州省气象台,贵州 贵阳 550002;
4. 贵州省气象局,贵州 贵阳 550002
Author(s):
GUO Xi1ZHAO Fou-xi2PENG Yu-xiang3ZHI Ya-jing1WANG Hua1WANG Juan1LIU Guo-qiang4*
1. Guizhou Meteorological Data Center,Guiyang 550002,China;
2. Guizhou Center for Disease Control and Prevention,Guiyang 550002,China;
3. Guizhou Provincial Meteorological Observatory,Guiyang 550002,China;
4. Guizhou Meteorological Bureau,Guiyang 550002,China
关键词:
气象因子脑卒中发病率注意力机制SCAD先验知识
Keywords:
meteorological factorsstroke incidenceattention mechanismsmoothly clipped absolute deviation (SCAD)prior knowledge
分类号:
TP181;R743.3;P49
DOI:
10.20165/j.cnki.ISSN1673-629X.2025.0005
摘要:
该文提出了基于平滑削边绝对偏离(SCAD)先验和注意力机制的脑卒中发病率预测模型。 通过深入研究,针对气象特征复杂多变等因素,设计了气象特征注意力模块,以提高模型的特征表达能力,并利用多头注意力机制,有效捕捉气象因子与脑卒中发病率的权重关联信息,降低了无用信息干扰,使得训练过程中更关注对脑卒中发病有显著影响的气象因子。 为了在本研究中数据规模有限的情况下,进一步提升模型的预测性能,该文结合了 SCAD 特征筛选方法和先验知识方法,设计了 SCAD 先验知识模块,帮助模型更快收敛,降低模型对数据规模的依赖。 该算法相比于多个对比模型在各项评价指标上都有所提升,其中相比于基线算法中表现最好的 Xgboost 模型,MSE 降低了 0. 064, R2 评分提高了 0. 054。 此外,对该算法进行了消融分析,验证了该算法设计的核心模块的作用,以及对模型性能提升的贡献。 实验结果表明,基于该方法设计的模型提高了脑卒中发病率的预测精度,适用于贵州山区脑卒中发病规律的目标识别。
Abstract:
An attention mechanism and Smoothly Clipped Absolute Deviation (SCAD) prior-based stroke incidence prediction model is developed. Through in-depth research,the meteorological feature attention module is designed to improve the feature expression ability of the model, and the multi - head attention mechanism is used to effectively capture the weight association information between meteorological factors and stroke incidence, reduce the interference of useless information, and make the training process pay more attention to meteorological factors that have a significant influence on stroke incidence. In order to further improve the prediction performance of the model in the case of limited data scale,we combine SCAD feature screening method and prior knowledge method to design SCAD prior knowledge module to help the model converge faster and reduce the model’s dependence on data scale. The proposed algorithm has improved in different evaluation metrics when compared to numerous comparative models. The MSE is decreased by 0. 064 for each of them when compared to the baseline algorithm’s top-performing Xgboost model,and the R2 score is raised by 0. 054 for each.The suggested algorithm is also subjected to an ablation analysis,which verifies the function of the core modules created for the algorithm in this chapter and its contribution to the enhancement of model performance. In conclusion, the proposed model can aid in stroke incidence prevention by properly predicting the stroke incidence based on meteorological data.
更新日期/Last Update: 2025-05-10