基于Storm的在线序列极限学习机的气象预测模型

发布时间：2018-02-14 23:52

本文关键词： Storm 极限学习机气象预测在线序列机器学习　出处：《计算机研究与发展》2017年08期 　论文类型：期刊论文

【摘要】：为提高气象预测精度,实时应对频发的局域气象灾害,拥有更高的处理海量数据的效率,提出了一种基于Storm的在线序列的极限学习机气象预测模型.该模型首先初始化多个在线极限学习机,当新批次的数据不断到达时,模型能够在训练结果的基础上继续学习新样本,并引入随机梯度下降法和误差权值调整方法,对新的预测结果进行误差反馈,实时更新误差权值参数,以提高模型预测准确率.另外,采用Storm流式处理框架对提出的算法模型进行并行化改进,以提高处理海量高维数据的能力.实验结果表明:该模型与基于Hadoop的并行极限学习机算法(parallel extreme learning machine,PELM)相比,具有更高的预测精度和优异的并行性能.
[Abstract]:In order to improve the precision of meteorological prediction and to deal with frequent local meteorological disasters in real time, it has higher efficiency of processing massive data. In this paper, a meteorological prediction model of extreme learning machine based on Storm is proposed. The model initializes several online extreme learning machines, and when the data of new batch arrive continuously, The model can continue to study new samples on the basis of the training results and introduce the stochastic gradient descent method and error weight adjustment method to carry out error feedback on the new prediction results and update the error weight parameters in real time. In order to improve the accuracy of model prediction, the proposed algorithm model is parallelized by using Storm flow processing framework. The experimental results show that the proposed model has higher prediction accuracy and better parallel performance than parallel extreme learning machine (PELM), a parallel extreme learning machine algorithm based on Hadoop.
【作者单位】：湘潭大学信息工程学院;智能计算与信息处理教育部重点实验室(湘潭大学);
【基金】：国家自然科学基金项目(61672495) 湖南省教育厅重点项目(16A208)~~
【分类号】：P45;TP181

【参考文献】