基于原型抽象和分类价值量的决策树算法*

为了克服现有决策树分类算法在大数据集上的有效性和可伸缩性的局限，提出一种新的基于粗糙集理论的决策树算法。首先提出基于代表性实例的原型抽象方法，该方法从原始数据集中抽取代表性实例组成抽象原型，可缩减实例数目和无关属性，从而使算法可以处理大数据集；然后提出属性分类价值量概念，并作为选择属性的启发式测度，该测度描述了属性对分类的贡献价值量的多少，侧重考虑了属性之间以及实例与分类之间的关系。实验表明，新算法比其他算法生成的决策树规模要小，准确率也有显著提高，在大数据集上尤为明显。

Decision tree algorithm using archetype abstraction and attribute classification value

In order to overcome the shortcomings of decision tree algorithm in large data sets, this paper proposed a novel decision tree algorithm based on rough set. The algorithm put forward a method based on representative instance for archetype abstraction, which extracted representative instances from original data set as abstraction archetype and decreased the number of instances and irrelevant attributes, hence it could deal with large data set. Simultaneity, the algorithm took attribute classification value as a heuristic measure for choosing attribute, which synthetically calculated contribution of an attribute for classification. It principally considered the dependency between attributes or relationship between instances and the classification. Mining experiments show that it can obtain higher accuracy and smaller size of decision tree than other algorithms, which make it more excellent for large data sets.