Abstract: For the objective demand of data mining technology on discretization and the characteristics of workshop production data,a discretization method of continuous attributes was established based on dynamic clustering.The objective function was established by using compatibility principle of decision-making system.The Density-Based Spatial Clustering of Applications with Noise (DBSCAN) was improved and a single dimension and multiple radius DBSCAN algorithm was proposed.The whole process of the discrete algorithm was elaborated.The experimental results proved that the discrete algorithm could keep the intrinsic compatibility of the decision-making system and possesses the merits such as fast speed,low memory occupancy rate,high level of automation and good generality,which was suitable for discretization of production data.

Key words: discretization, production scheduling, data mining, density-based spatial clustering of applications with noise

摘要： 针对车间生产数据的特点及数据挖掘技术对离散处理的客观需求,建立一种基于动态聚类的连续值离散化方法,并利用决策系统的相容性原则建立目标函数。对基于密度的聚类算法进行改进,提出一种单维度多半径聚类算法。将该聚类算法应用于离散处理,阐述了基于动态聚类离散算法的整体过程。实验数据表明,该离散算法可以保持决策系统原有的相容度,具有速度快、内存占用率低和自动化程度高等优点,且具有良好的通用性,适用于生产数据的离散处理。

关键词: 离散化, 生产调度, 数据挖掘, 基于密度的聚类算法