摘要： 为提高Web数据表识别的准确性，提出一种基于支持向量机与混合核函数的数据表识别方法。给出表格的结构特征、内容特征以及行(列)相似特征，将多项式核函数和线性核函数组成混合核函数，利用其进行Web数据表的自动识别。实验结果表明，该方法在7个站点上，准确率和召回率的平均值为95.14%和95.69%。

关键词: Web页面, 数据表, 特征抽取, 支持向量机, 核函数

Abstract: In order to improve the identification accuracy of Web data table, this paper proposes an identification method based on Support Vector Machine(SVM) and mixed kernel function. This paper gives the structural features, content features and row(column) similarity features of the table, and takes mixed kernel function constructed by a polynomial kernel function and a linear kernel function, automatically recognizes the Web meaningful tables. Experimental result shows that the average precision rate and recall rate of this method are 95.14% and 95.69% in seven sites.

Key words: Web page, data table, feature extraction, Support Vector Machine(SVM), kernel function

中图分类号: 

• TP391