摘要
粗糙集属性约简是一个典型的NP-hard问题。提出了一种基于子集类蚁群模型的属性相对约简算法,该算法采用转移概率对每个属性随机搜索,直到获得一个分类能力与决策属性分类能力一致的属性子集。提出的基于信息素变异的蚁群算法,不仅提高了解的质量,而且有效避免了早熟收敛。106组病例数据的实验结果表明,该算法能够发现较好的决策表相对约简与决策规则。
Reduction in rough set theory is a typical NP-hard problem. A new algorithm for relative reduction which is based on Subset ant colony algorithm was proposed. By using the strategy of diversion probability, every attribute is randomly searched by ants until an attribute subset is obtained that has the same discerning capability with the decision attribute. The presented ACO algorithm based on pheromone mutation not only improves the solution quality but also avoids stagnation. The experiment with 106 illness cases shows that the algorithm can discover better relative reduction and decision rules.
出处
《计算机科学》
CSCD
北大核心
2008年第11期147-150,共4页
Computer Science
基金
国家自然科学基金项目(No.60573074)
忻州师范学院基金资助项目
关键词
粗糙集
子集类蚁群算法
属性约简
信息素变异
Rough set, Subset ant colony algorithm, Reduction, Pheromone mutation
