摘要
为了对烟草病毒病的病情指数进行建模研究,采用了一种正则极限学习机算法:通过引入惩罚因子来权衡结构风险与经验风险的大小,进一步增强网络的泛化性.针对烟草病毒病的众多影响因素,采用灰色关联度算法选取主要影响因子.使用某地1984-1995年病情资料、相关虫情和气象资料,经过数据挖掘、建模仿真,将正则极限学习机应用于烟草病毒病预测中,效果较好,对烟草病毒病的防治具有指导意义.
To model the disease index of tobacco virus diseases, a regular extreme learning machine (RELM) algorithm was adopted. By introducing the penalty factor to balance the structural risk and em- pirical risk, the generalization of the network was further enhanced. Considering the influencing factors of tobacco virus diseases, gray relational degree algorithm was used to select the main impact factors. A group data of disease, related pests and meteorological conditions, which was coming from a certain place from 1984 to 1995, was used to verify the effectiveness of the proposed RELM method. The group data was processed by data mining and modeling, then RELM was applied to predict the tobacco virus disea- ses. The RELM method demonstrated better performances than traditional ELM and had guiding signifi- cance for the control of tobacco virus diseases.
出处
《郑州大学学报(理学版)》
CAS
北大核心
2013年第4期58-62,共5页
Journal of Zhengzhou University:Natural Science Edition
基金
国家自然科学基金资助项目
编号60905039
F030507
关键词
烟草病毒病
正则极限学习机
灰色关联度
建模
tobacco virus disease
regular extreme learning machine
gray relational degree algorithm
modeling