基于SVM用户建模的核函数选择研究  

Selection of Kernel Function in Support Vector Machine Based User Modeling

作　　者：方若宇[1,2] 张琼[1] 许慰玲[3] 张虹[2] 

机构地区：[1]汕头大学工学院,广东汕头515063 [2]汕头大学理学院,广东汕头515063 [3]仲恺农业工程学院信息学院,广东广州510225

出　　处：《微计算机信息》2011年第8期241-243,共3页Control & Automation

基　　金：汕头大学青年基金;广东省轻工装备产学研示范基地开放课题的支持

摘　　要：用户建模是从用户偏好数据中建立用户偏好模型的过程,用户偏好数据具有系统运行初期的稀疏性和非线形的特点。支持向量机(Support Vector Machine,简称SVM)具有小样本学习、非线形处理的能力,是合适的用户建模工具。SVM的非线形处理能力主要依赖于核函数,采用不同的核函数进行建模对模型的预测效果有重大影响。本文重点研究核函数的选择对基于SVM建模方法的影响,从中选取了表现较优的小波核函数,构建性能突出的SVM进行用户建模。实验证明该建模方法可以有效地从小样本数据中学习用户偏好信息,建立反映用户真实偏好的用户模型。User modeling is the process of modeling user preference from selecting user preference data which is characteristic of sparseness and non-linear in system first operating period.Support vector machine,which is capable of small sample learning and non-linear processing,is a suitable way of user modeling.Non-linear processing of support vector machine mainly depends on kernel function.Therefore,selecting deferent kernel functions has significant impact on the effectiveness of user modeling.This paper focuses on the impacts of kernel function selection on support vector machine based user modeling and build support vector machine with better performance via selecting better kernel function.The experiment shows that the approach is capable of rapid prototyping in condition of small sample sets and extracting the real user preference.

关 键 词：用户建模 支持向量机 核函数 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]