摘要
属性约简在粗糙集理论研究中一直占据重要位置。为了能够更加快速有效的获得决策表中属性的最优约简。提出了一种新的启发信息遗传算法的粗糙集属性约简算法。引入属性频率作为启发式信息构造适应度函数，相比于传统矩阵方法，减少了大量矩阵操作。在交叉操作时，基于属性重要度的特性，引入判别属性相似度这一操作，父代相似个体不进行交叉，避免了不必要的个体交叉。实验结果表明，该算法比传统方法更准确的获得关键属性，且迭代的次数更少，能更有效地约简属性。
Attribute reduction is a key problem for the rough set theory. In order to be more effective to obtain the minimal attri- bute reduction, the paper put forward a new attribute reduction algorithm based on GA, In the fitness selection, proposing the attri- bute frequency as the heuristic information, compared to the traditional matrix method, it has reduced the amount of operation. In the crossover operation, introducing the distinguishing attribute similarity which based on the importance of attribute in the cross- over operation, so similar parents do not cross, so as to avoid unnecessary individual cross. The experimental results show that, the algorithm can not only obtain the excellent properties more accurately, but also can reduce the number of iterations, it can quickly and effectively to attribute reduction.
出处
《电脑知识与技术》
2015年第3期281-285,共5页
Computer Knowledge and Technology
基金
江苏省自然科学基金（BK2012209）
苏州市工业应用基础研究项目（SYG201409）
关键词
粗糙集
属性约简
遗传算法
属性频率
属性相似度
rough set
attribute reduction
genetic algorithm
attribute frequency
attribute similarity
