基于蜂群优化的粗糙集属性约简

为了获得决策表中更好的属性约简,提出一种信息增益引导的蜂群优化算法;该算法以属性的信息熵为基础构造条件属性与决策属性间的互信息,用待选条件属性引起的信息增益作为引导蜜蜂搜索的启发信息,最终求得属性约简集;对UCI数据库多个数据集的测试结果表明,与其它基于群智能的属性约简算法相比,该算法获得最小属性约简的机率提高到90%以上,同时较对比算法的计算时间少耗费至少10%。

Rough Set Attribute Reduction based on Bee Colony Optimization

In order to obtain better attribute reduction in decision tables,a bee colony optimization algorithm for attribute reduction based on information gain is proposed.Information gain is constructed based on the mutual information between selected conditional attributes and decision attributes by information entropy of attributes.The algorithm dynamically calculates heuristic information based on information gain to guide search.The results demonstrate that proposed algorithm can get better results than other intelligent swarm algorithms for attribute reduction in term of both solution quality and computational effort,reduce time consumption around 10%,and improve probabilistic of get the least reduction of attributes nearly 90%.