摘要
将贝叶斯证据框架应用于最小二乘支持向量机模型参数的选择,建立起时间序列的非线性预测模型.在推断的第一层,选择模型的参数,在推断的第二层,选择模型超参数,第三层,选择模型的核参数,并选择相关输入变量.该模型应用于加州电力市场现货价格的预测,取得了较好的效果.
In this paper, we applied the Bayesian evidence framework to least squared support vector machine （LS- SVM） regression in order to infer nonlinear models for predicting a time series. On the first level of inference, model parameters were selected and on the second level the hyperparameters were selected. The kernel parameter were tuned on the third level framework, and on this level the relevant inputs were selected. We demonstrated the utility of this approach on prices prediction of California electricity market.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2007年第5期142-146,共5页
Systems Engineering-Theory & Practice
