分类规则的蚁群挖掘算法

Ant Colony Mining Algorithm under the Classification Rule

提出了一种分类规则的蚁群挖掘算法。算法首先对所有的连续属性值离散化,得到相应的离散属性,然后让各只蚂蚁按照某种策略选择相关属性,对所选属性再选择理想的属性值,循环地构造单个规则,接着更新训练集,最终形成各类规则集。最后用新规则约简算法进行约简操作。对两个公用数据的实验及其与Ant-Miner和C4.5的对比表明,算法能够发现更好的分类规则。实验同时表明该算法是有效的。

The paper proposes an ant colony mining algorithm under the classification rule. Firstly, it obtains corresponding discrete attributes through discretizing all the continuous attribute values, then lets each ant choose the relevant attributes in accordance with a certain strategy and select the desired attribute value of the selected attributes, to circularly construct the single rules and update the training set, and generate various types of rule sets. Finally, it reduces the rules with new Reduction rules algorithm. Experiments on two public data sets and the comparison with Ant-Miner and Contrast with C4.5 show that the algorithm can discover better classification rule and the algorithm is effective.