基于主成分分析的PSO-BP算法对肥胖水平预测研究  

Research on obesity level prediction by the PSO-BP algorithm based on principal component analysis

作　　者：邱麒添 Qiu Qitian(School of Mathematics and Systems Science,Guangdong Polytechnic Normal University,Guangzhou 510665,China)

机构地区：[1]广东技术师范大学数学与系统科学学院,广州510665

出　　处：《现代计算机》2024年第19期13-18,共6页Modern Computer

摘　　要：提出一种主成分分析法(PCA)与粒子群优化BP神经网络算法(PSO-BP)相结合的肥胖水平预测模型。通过主成分分析法对16个输入变量进行降维,提取出11个综合变量作为BP神经网络的输入,利用PSO算法优化BP神经网络的权值与阈值,进一步提升网络训练的能力。实验结果表明,基于主成分分析的PSO-BP算法不仅简化了模型结构,而且在更短的时间内实现了对肥胖水平更高的分类预测精度。这一研究为个体肥胖水平的评估提供科学的依据,具有重要的研究意义与应用价值。An obesity level prediction model combining principal component analysis(PCA)and particle swarm optimization BP neural network algorithm(PSO-BP)is proposed.Dimension reduction of the 16 input variables are performed by principal component analysis,and 11 comprehensive variables which represented principal components are extracted as the input of BP neural network,optimize the weights and thresholds of network by using the PSO algorithm,further improve the ability of network training.Experimental results show that the PSO-BP algorithm based on PCA achieves higher classification prediction power at a shorter time cost.This study provides a scientific basis for the evaluation of individual obesity level,and has important research significance and application value.

关 键 词：肥胖水平 主成分分析法 PSO-BP算法 分类预测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] R589.2[自动化与计算机技术—控制科学与工程]