一种基于支持向量机的模糊施肥预测方法研究  

A NEW APPROACH FOR FUZZY FERTILIZATION FORECAST BASED ON SUPPORT VECTOR MECHANISM

作　　者：李淼[1] 方德洲[1] 张建[1] 徐大庆[1] 张强[1] 

机构地区：[1]中国科学院合肥智能机械研究所,安徽合肥230031

出　　处：《计算机应用与软件》2009年第2期50-52,共3页Computer Applications and Software

基　　金：中科院知识创新工程重要方向项目(KGCX2-SW-511);国家科技攻关计划(2005BA113A03)

摘　　要：提出了一种利用支持向量机(SVM)学习算法提取模糊规则进而实现施肥预测的方法。对于农业施肥中常用的正交实验法,由于其数据均衡分散使得曲线拟合的回归预测方法效果不佳。提出了一种利用SVM学习样本数据,再利用隶属度来提取模糊规则,通过阈值和可信度来控制规则的激活和准确性的预测方法,这一方法不仅避免了回归预测所产生的误差,并且模糊规则更具有实际意义,从而大大提高了知识获取的能力。In this paper it has proposed a method which makes use of the teanfing algorithm of support vector machine （ SVM ） to extract fuzzy rules and then realizes the fertilization forecast. Orthogonal experiment is used in agriculture fertilization widely, but the effect of curve fitting regressive forecast method is not very＂ well due to the even dispersing of its sample data. The method proposed in this paper is based on SVM learning sample data. It is a forecast method which controls the activation of rules and the accuracy by threshold value and credibility where the fuzzy rules are extracted according to the membership degree of each support vector projection. Not only the method has avoided the error produced by regressive foreeast, but also the fuzzy rules have the actual significance more＇, so the ability of knowledge acquisition has been improved greatly.

关 键 词：支持向量机 模糊集合 规则提取 阈值 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] O212[自动化与计算机技术—控制科学与工程]