基于决策树和链接相似的Deep Web查询接口判定*

针对现有Deep Web查询接口判定方法误判较多、无法有效区分搜索引擎类接口的不足，提出了基于决策树和链接相似的Deep Web查询接口判定方法。该方法利用信息增益率选取重要属性，并构建决策树对接口表单进行预判定，识别特征较为明显的接口；然后利用基于链接相似的判定方法对未识别出的接口进行二次判定，准确识别真正查询接口，排除搜索引擎类接口。结果表明，该方法能有效区分搜索引擎类接口，提高了分类的准确率和查全率。

Deep Web query interface identification based on decision tree and link-similar

In order to solve the problems existed in the traditional method that Deep Web query interfaces are more false positives and search engine class interface can not be effectively distinguished, this paper proposed a Deep Web query interface identification method based on decision tree and link-similar. This method used attribute information gain ratio as selection level, built a decision tree to pre-determine the form of the interfaces to identify the most interfaces which had some distinct features, and then used a new method based on link-similar to identify these unidentified again, distinguishing between Deep Web query interface and the interface of search engines. The result of experiment shows that it can enhance the accuracy and proves that it is better than the traditional methods.