摘要： 针对覆盖算法中识别精度与泛化能力之间的一对矛盾，结合佳点集遗传算法思想，提出佳点集遗传覆盖算法。通过在覆盖种群中引入竞争，淘汰不好的覆盖，保留较优的覆盖，减少了覆盖个数和拒识样本个数，从而提高了整体覆盖种群的分类能力。与Lib-SVM、领域覆盖、交叉覆盖的对比实验证明了该算法具有良好的分类识别精度与泛化能力。

关键词: 佳点集遗传算法, 机器学习, 领域覆盖

Abstract: Aiming at solving the conflict between validity and extensibility of the covering algorithm, this paper combines good point-set genetic algorithm with covering algorithm, and presents a good point-set genetic covering algorithm. The algorithm brings competition in the population of cover sets, eliminates the poor cover and retains the better cover, signally reduces the number of coverage and the number of samples rejection, so it raises the classification capacity of the total population. By comparing with Lib-SVM and neighborhood covering and alternative covering algorithm, experimental results prove that this algorithm has a good validity and extensibility.

Key words: good point-set genetic algorithm, machine learning, neighborhood covering

中图分类号: 

• TP301.6