文献类型：期刊文献

中文题名：基于在线序列-极限学习机的干旱预测

英文题名：Drought Prediction Based on Online Sequential Extreme Learning Machine

作者：刘振男 周靖楠

第一作者：刘振男

机构：[1]贵州理工学院;[2]河海大学

第一机构：贵州理工学院

年份：2018

卷号：39

期号：8

起止页码：84-87

中文期刊名：人民珠江

外文期刊名：Pearl River

基金：贵州省科技厅基金项目(黔科合LH字[2016]7096)

语种：中文

中文关键词：极限学习机;在线序列;干旱;预测因子

外文关键词：Extreme Learning Machine;online sequential;drought;forecasting factor

摘要：极限学习机在干旱预测时,通常将作为预测因子的历史数据固化的导入到模型中进行训练,而忽略了不同阶段产生的数据在模型训练中的作用和效果。因此,基于在线更新理论构建了在线序列-极限学习机预测模型,该模型在参数训练更新时,预测因子数据是按不同批次逐步导入到模型进行训练,大大降低了计算机资源占用率,且选用标准降水指数作为干旱评价指标对研究区域进行了预测。结果表明:在线序列-极限学习机较极限学习机的预测精度与稳定性有了大幅度的提高。
The historical data as the predictors are usually imported into a model of extreme learning machine for training in drought prediction,while ignoring the effect by the different stages of data for training in the model,therefore,online sequence extreme learning machine is built based on the online renewal theory,the parameters in the model are updated with the predictor data are imported into the model step by step,and the computer resource is reduced.The standard precipitation index is selected as evaluation index for drought.The results show that the precision and stability of online sequence extreme learning machine are better than extreme learning machine significantly.