基于模式分类的自学习生产调度知识库方法

Constructing Production Scheduling Knowledge System ——A Machine Learning Method Based on Pattern Classification

对一种简化的调度问题进行了分析 ,说明了在一定的条件下 ,对某一范围的调度任务 ,的确存在着一种不变的模式 .文中对任务中不同的子任务对 (“订单对”)给出了一个指标 ,并以此指标来判断该子任务对在一个成功的生产调度方案中的先后顺序 ,是否在另一个类似的任务的成功生产调方案中能得到保持 .基于这一结果 。

A simplified scheduling problem is studied in this paper, with a conclusion that under certain conditions, there exists an invariable pattern among the successful solutions of the scheduling tasks within a certain bound. An index is given to different pairs of subtasks(i.e., ordering\|pairs) in a task to estimate whether the order of the two subtasks in a certain ordering\|pair will be kept in other similar tasks. Based on those results, an algorithm is proposed to execute machine learning on the empirical scheduling plans.