摘要
根据不同温度下的玻璃熔窑大碹横向膨胀率的实测数据集,应用基于粒子群算法寻优的支持向量回归结合留一交叉验证的方法对玻璃熔窑大碹横向膨胀率进行了建模和预测,并将其预测结果与最小二乘回归进行了比较。结果表明:留一交叉验证法的支持向量回归预测的均方根误差、平均绝对误差和平均绝对百分误差均为最小。因此,支持向量回归是一种预测玻璃熔窑大碹横向膨胀率的有效方法。
The support vector regression approach combined with particle swarm optimization for its parameter optimization is proposed to conduct leave-one-out cross validation for the horizontal expansion ratio of glass furnace big spin under different temperature, and the best prediction performances were provided by it. The results strongly support that the generalization ability of the LOOCV test of SVR surpasses that of least square regression. These suggest that SVR ,may be a promising and practical technique to accurately estimate the expansion ratio of glass furnace big spin.
出处
《玻璃》
2011年第12期18-22,共5页
Glass
关键词
玻璃熔窑
大碹
横向膨胀率
支持向量回归
粒子群优化算法
留一交叉验证法
预测
glass furnace
big spin
horizontal expansion ratio
support vector regression
particle swarm optimization
leave one out cross validation
prediction
