摘要： 在覆盖算法中，识别精度与泛化能力之间存在矛盾。为此，结合粒子群优化(PSO)具有的全局搜索能力，提出一种PSO覆盖算法。将领域覆盖算法中每一类样本形成的一组覆盖转化为粒子群，并在迭代过程中搜索出较好的覆盖粒子，从而得到一组个数较少且分类效果较好的覆盖。实验结果表明，该算法具有较高的分类识别精度及较优的泛化能力。

关键词: 粒子群优化, 机器学习, 覆盖算法, 全局搜索, 泛化

Abstract: Aiming at the conflict between recognition accuracy and generalization ability. Considering the stronger search ability of Particle Swarm Optimization(PSO), this paper introduces a PSO covering algorithm. The algorithm converts a group of coverings of each type of sample given with covering algorithm into a particle swarm, and searches for a better covering particle in the iteration process. It gets a set number of coverings, which is less but with better classification performance. Experimental results prove this algorithm has better classification accuracy and generalization capability.

Key words: Particle Swarm Optimization(PSO), machine learning, covering algorithm, global search, generalization

中图分类号: 

• TP311.52