摘要
在Ant-Miner算法基础上提出了一种利用蚁群算法解决分类规则挖掘的算法(ACR),设计了合理的蚂蚁选择属性及属性分区的概率公式,并对规则质量的衡量等策略进行改进,可以较好地挖掘分类规则.在标准数据集上通过与Ant-Miner算法和经典的基于决策树的C 4.5算法比较,ACR在挖掘分类规则的简单性、正确率上有较好的表现.
Based on the Ant-Miner algorithm, the paper presents a new algorithm for classification rule mining problem in which ant colony algorithm is used. The algorithm uses reasonable probability formula for ant's attributes and attributes values and employs the strategy for rule quality measurement. Therefore, ACR could better discover the classification rules. Compared with Ant-Miner algorithm and the classical C 4.5 algorithm based on decision tree, the experimental results on several benchmark datasets show that ACR can discover classification rules with better simplicity and quality.
出处
《江南大学学报(自然科学版)》
CAS
2008年第5期511-515,共5页
Joural of Jiangnan University (Natural Science Edition)
基金
国家自然科学基金项目(60773206/F020106)
关键词
蚁群算法
分类问题
规则发现
数据挖掘
ant colony algorithm
classification
rule discovering
data mining