Abstract: Pattern-oriented workflow generation has utilized AI planning technologies and knowledge for automatic workflow construction.The quality of process pattern has essential impact on workflow generation result.This paper proposes an approach for pattern knowledge optimization.Based on expert modeling result，scenario classifier is constructed.Process pattern can be improved by classifier training result on planning history，which enable workflow generation more reliable.

Key words: workflow generation, process pattern, scenario classifier training

摘要： 基于模式的工作流生成利用规划技术和业务知识实现了工作流自动创建。流程模式知识的优劣直接影响到工作流生成结果的正确性。提出了一个针对流程模式的知识优化方法。在人工建模的基础上，针对流程模式知识构建场景分类器。利用流程规划运行的历史数据，通过机器学习对流程模式进行调整与优化，保证了工作流创建的正确性。

关键词: 工作流生成, 流程模式, 场景分类器训练