一种求解分类问题的自适应人工蜂群算法

Adaptive artificial bee colony algorithm for classification problem

考虑到分类算法学习到的分类器的可理解性,提出一种求解分类问题的自适应人工蜂群算法——A_ABC,该算法生成一组可理解的分类规则。在基于规则的分类方法中,采用合适的规则评价函数能够提高分类算法的性能,A_ABC算法能够针对不同数据集自适应选取相适应的规则评价函数,同时能够有效处理连续类型的属性和离散类型的属性。最后,在多个公用的真实数据集上,将A_ABC算法与相关算法进行了比较,结果表明A_ABC算法能够更加有效地解决分类问题。

Appropriate rule evaluation functions are important to improve the performance of classification algorithm based on rules.In order to obtain the comprehensible classification rules,an adaptive artificial bee colony algorithm,called A_ABC algorithm,is proposed.The A_ABC algorithm can adaptively select a appropriate rule evaluation function for the given data,and effectively process both discrete attributes and continuous attributes.The proposed A_ABC algorithm is evaluated by experiment using different standard real datasets,and compared with existing classification algorithms.Results show that the A_ABC algorithm can solve classification problems more effectively than the existing algorithms.