查询结果:   郑光键.基于流程挖掘的业务流程模块推荐方法研究[J].计算机应用与软件,2018,35(6):181 - 189.
中文标题
基于流程挖掘的业务流程模块推荐方法研究
发表栏目
人工智能与识别
摘要点击数
675
英文标题
RESEARCH ON THE METHOD OF BUSINESS PROCESS MODULE RECOMMENDATION BASED ON PROCESS MINING
作 者
郑光键 Zheng Guangjian
作者单位
复旦大学软件学院 上海 200433     
英文单位
School of software, Fudan University, Shanghai 200433, China     
关键词
流程模块推荐 主题模型 神经网络 绩效预测
Keywords
Process module recommendation Topic model Neural network Performance prediction
基金项目
作者资料
郑光键,硕士生,主研领域:商务智能,数据挖掘。 。
文章摘要
业务流程设计在流程管理中起到了非常关键的作用,但同时也比较耗时费力。为了在设计过程中给出合理的业务流程模块推荐,使得流程绩效最大化,提出一种基于流程挖掘技术的业务流程模块推荐方法。首先基于流程日志建模,提取活动特征后将流程模块化分解。再使用改进的基于最优主题结构的隐狄克雷分布LDA(Latent Dirchlet Allocation)主题模型提取算法提取流程模块主题模型。同时构建业务流程整体特征后,基于小批量梯度下降法的BP神经网络算法训练流程绩效预测模型。将用户建模需求做所需功能主题提取后,基于K近邻思想生成功能主题模块集合,以最大化流程绩效为目标对流程模块进行动态组合推荐。实验表明,基于流程挖掘的业务流程模块推荐方法能够提高流程模块推荐的准确性,也更加符合流程模块推荐中的语义性需求与功能主题的特性。
Abstract
Business process design plays a key role in process management, but it is also time-consuming and labor-intensive.In order to give reasonable business process module recommendations in the design process and maximize process performance, a business process module recommendation method based on process mining technology was proposed. Based on process log modeling, the process features were extracted and the process was modularly decomposed.Then we used the improved Latent Dirchlet Allocation (LDA) topic model extraction algorithm based on the optimal topic structure to extract the process model topic model. Meanwhile, the BP neural network algorithm based on small-batch gradient descent method was used to train process performance prediction model after constructing the overall business process characteristics. After the user modelling needs were extracted as the required functional topics, a functional theme module set was generated based on the K-nearest neighbor idea, and the process module was dynamically combined and recommended with the goal of maximizing the process performance. Experiments show that the business process module recommendation method based on process mining improves the accuracy of the process module recommendation, and is more in line with the semantic requirements and functional features of the process module recommendation.
下载PDF全文