摘要
多元自适应回归样条建模中,样本数据最小步长和端点的设置会影响建模精度。提出了应用粒子群算法优化这两个参数的方法,以预测样本均方差为适应度函数,通过优化最小步长和端点位置调整采样点选取。实例应用表明,PSO-MARS方法能提高预测精度。
In the multi-adaptive regression spline modeling process,the setting of both sampling minimum step size and endpointmay influence the precision of modeling.To optimize the two parameters,Particle Swarm Optimization(PSO)method is applied to estimate the Mean Square Error(MSE).The MSE is taken as the fitness function to optimize the minimum step size and endpoint by adjusting the sampling position.Application results indicate that themethod can improve the modeling accuracy.
出处
《长春工业大学学报》
CAS
2017年第5期459-463,共5页
Journal of Changchun University of Technology
基金
吉林省科技发展计划基金资助项目(20150203003SF)
关键词
粒子群优化
最小步长
端点
交叉验证
Particle Swarm Optimization(PSO)
minimum step size
endpoind
cross-validation
