基于蚁群遗传算法的属性约简

Attribute Reduction Based on Ant Colony Genetic Algorithm

针对普通蚁群算法在属性约简中求解最小约简存在局部最优、迭代次数多、收敛慢的问题,将复制、交叉、变异这些遗传算子引入蚁群算法中,改进蚂蚁的产生方式和蚂蚁构造可行解的过程,提高算法的收敛速度和全局搜索能力。算法在加州大学机器学习数据库中的数据集的测试结果表明,该算法能快速有效地求解属性约简,能够找到最小约简集。

As for the ordinary ant colony algorithm for attribute reduction, which has the problems such as local minima, many it- erations and slow convergence, this paper proposes the ant colony genetic algorithm that takes copy, crossover and mutation of ge- netic operators to ant colony algorithm, which can improve the generation of ants and the process of the feasible solution, to im- prove global search capability. The algorithm is validated on data sets of UCI machine learning database from the University of California. The results show that the algorithm can quickly and efficiently solve the attribute reduction, to be able to find the min- imal reduction set.