PVS-PSO-SVR协同模型及实证分析

PVS-PSO-SVR cooperative model and its empirical analysis

针对高维随机变量信息冗余以及主成分分析降维的缺陷,用主变量筛选法对高维随机变量降维,利用提取的主变量建立了支持向量回归机(SVR)模型.对于模型的参数,利用了改进的粒子群算法进行优化选择.构建出主变量筛选(PVS)、粒子群优化(PSO)和SVR的协同模型,并用于葡萄酒的质量预测.实验证明PVS-PSO-SVR协同模型与已有的3种模型(N-CV-SVR模型、PCA-CV-SVR模型,PVS-CV-SVR模型)相比,检查误差有较大的改善,表明PVS-PSO-SVR协同模型泛化能力强、预测结果更有效.

In response to the redundancy of high-dimensional random variable information and the shortcomings of principal component analysis in dimensionality reduction,the principal variable screening method was used to reduce the dimensionality of high-dimensional random variables.A support vector regression machine model was established using the extracted principal variables For the parameters of the model,an improved particle swarm optimization algorithm was used for optimization selection Construct a collaborative model of Principal Variable Screening(PVS),Particle Swarm Optimization(PSO),and Support Vector Regression(SVR)for wine quality prediction.Demonstration experiments shown that the PVS-PSO-SVR collaborative model has significantly improved inspection errors compared to the existing three models(N-CV-SVR model,PCA-CV-SVR model,PVS-CV-SVR model),which indicates that the collaborative model of principal variable selection,particle swarm optimization,and support vector regression has a stronger generalization ability and more effective prediction results.