Abstract: In order to solve the problem of data mining and the discretization of continuous attributes in the field of machine learning, an improved adaptive discrete particle swarm optimization algorithm is proposed. This method treats the discrete particle swarm as a breakpoint set of continuous attributes. It also minimizes breakpoint subset through the interaction of particles, combined with simulated annealing algorithm as a partial search strategy for particles, enriching the particle swarm and enhancing the ability to look for the whole optimal solution. In addition, the consistency of decision table is measured according to the dependence of decision attribute in the rough set theory on condition attribute, achieving the goal of continuous attributes discretization. Finally the performance of this algorithm is tested through multiple sets of data and compared with other algorithms through experiments. As the results show, this algorithm is effective.

Key words: discrete particle swarm, simulated annealing, rough set, continuous attributes discretization

摘要： 为了解决数据挖掘和机器学习领域中连续属性离散化问题，提出一种改进的自适应离散粒子群优化算法。将连续属性的断点集合作为离散粒子群，通过粒子间的相互作用最小化断点子集，同时引入模拟退火算法作为局部搜索策略，提高了粒子群的多样性和寻找全局最优解的能力。利用粗糙集理论中决策属性对条件属性的依赖度来衡量决策表的一致性，从而达到连续属性离散化的目的，最后采用多组数据对此算法的性能进行了检验，并与其他算法做了对比实验，实验结果表明此算法是有效的。

关键词: 离散粒子群, 模拟退火, 粗糙集, 连续属性离散化