A support vector machine-based algorithm for mining the knowledge hidden in inconsistencies

A support vector machine-based algorithm is presented for mining the kinds of information hidden in inconsistent examples, i.e., deciding whether the inconsistencies are caused by mistakes, errors, or missing attributes. When the amount of inconsistent examples is less than the threshold, the inconsistencies are caused by mistakes. When the inconsistent examples belong to the concepts, which are not close to each other or the inconsistent examples are not on the boundary of two classes, the inconsistencies are caused by missing attributes, and some condition attributes which can be added to eliminate inconsistencies are given. When the inconsistent examples are support vectors, they may be caused by errors or missing attributes, the measures of reducing the errors should be taken firstly.