不完备日志的过程挖掘算法  

Algorithm of Incomplete Log's Processing Mining

作　　者：易科[1] 叶剑虹[1] 

机构地区：[1]华侨大学计算机科学与技术学院,福建厦门361021

出　　处：《小型微型计算机系统》2017年第11期2541-2546,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61170028;61573158)资助;华侨大学科技创新团队和领军人才支持计划项目(2014KJTD13)资助;华侨大学基本科研业务费专项基金项目(JB-ZR1130)资助;华侨大学研究生科研创新能力培育计划项目(1511414011)资助

摘　　要：算法能够在日志不完备的情况下,首先构造初始依赖图,根据依赖图所需满足的消极边约束条件,直接删除禁止的边,对于消极路径约束,需要针对约束关系进行分类以删除因果值最小的边.对于积极边约束,可直接添加所需的边,但对于积极路径约束,还需进一步计算出权值最优的路径以添加到模型中.为使构造出的网结构更吻合原始模型,随后还通过绑定操作生成一个支持原有日志轨迹的因果网.算法通过对消极的约束关系分类,减少了对消极约束的处理,同时在积极约束上采用权值更新最优算法提高效率.绑定操作可以进一步帮助我们约束网模型的行为,减少非原始日志轨迹的生成.整个挖掘算法,比已有的算法在效率上有显著提升,同时新增了一些对复杂约束条件的处理能力.In this paper,we can discover causal net with constraints in an incomplete log. A dependency graph can be constructed firstly,then the dependency graph remove directly forbidden edge on the basis of negative edge constraints,with regard to negative path constraints,the edge with minimal causal score will be removed by classifying constraint relationship. The required edge can be added to dependency graph directly with regard to positive edge constraints,but for positive path constraints,the path with optimal weight should be calculated out and be added to model. At last,in order to make constructed network fit original model,a casual net that supports original log traces can be built by binding operation. Our algorithms can reduce the steps in the operation of negative constraint,and the effective of process is improved in the operation of satisfying positive constraint by weight-best updating method. The model is restricted its behaviors by the binding operation algorithm,which can help us to avoid the state explosion problem. Some experiments appropriately prove that. The efficiency of process mining is improved and some complicated constraints condition can also be handled.

关 键 词：依赖图 因果网模型 消极约束 积极约束 日志 挖掘 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]