基于过程挖掘的网络化软件运行时行为分析方法

The advantage of process mining technology is to construct an important bridge between knowledge discovery and business process modeling, analysis and optimization. Based on this emerging technology, we use special viewponit to survey runtime behavior of networked software, which aims at to explore an efficient analysis method. The process mining technology is used to discover service behaviors and extract them into behavior model, and discover relatively invariable mode from interaction and coordination of multiple concurrence cases as their relationship. Multi-view and multi-level 3D description method is presented with the control flow as center. This method uses Interactive Behavior oriented Petri Net (IBoPN) to describe software behaviors,and the section of interaction and coordination is described with "synchronous block", which can efficiently express complex relationship of data objects and comple correspondence of multiple concurrence cases. The preconditions about data object operation, resource allocation and context environment are transformed into the constraints for Petri net. Thus, the problem of data-aware interactive behavior, which is very complicated originally, is simplfied into the problem of conformance checking of Petri net. The "replay" technology of process mining is adopted to detect and quantify the different between current events and behavior models, discover the bottleneck that restricts runtime behaviors, predict subsequent software behaviors, and provide credible evidences for improving software performance continuously.

过程挖掘技术的优势在于能在知识发现和业务过程建模、分析及优化之间搭建起重要桥梁.基于该新兴的技术,我们用独特视角审视网络化软件运行时行为,旨在探索一套有效的分析方法.提出用过程挖掘技术发现服务行为并将其提炼成行为模型,发现多个并行案例交互与协同时的相对不变模式,作为行为模型间的关联关系;提出以控制流为中心的多视角、多层次的3D描述方法,该方法用面向交互行为的Petri网描述软件行为,其中的交互与协同部分采用"同步块"描述,可有效表达数据对象间复杂关系及多个并行案例间的复杂通信;在交互行为分析时,数据对象操作、资源分配、上下文环境等前置条件被转换成对Petri网的约束,原本复杂的数据感知的交互行为分析问题被简化为对Petri网的符合性检查问题.利用过程挖掘的"再现"技术,可有效检测和量化当前事件和行为模型间的差异,发现制约运行时行为的瓶颈,预测后续软件行为,并为持续提升软件性能提供可靠依据.

为了从开放和动态的环境中把握网络化软件的可信性，需要从网络化软件运行时的交互行为这个全新的视角对其进行分析，过程挖掘技术是满足这一要求的有效手段。. 在前期初步提出软件行为描述、软件行为动态建模的基础上，本课题完成了如下工作：提出了利用过程挖掘技术构建网络化软件行为动态模型的方法，并研制了相应的可视化建模工具；提出网络化软件行为描述的形式化方法和约束规则，从而完善过程模型；针对松散聚合的软件实体之间的交互行为呈现出涌现行为的特征，提出了多视角、多层次的行为分析方法并实现相应支持工具；进一步地，对开放和动态环境下网络化软件的交互行为传播、牵制控制问题展开了研究，分析了影响异常行为传播的多种因素，结合原型系统实验揭示了异常行为传播的规律，提出了异常行为的牵制控制方法。. 已发表学术论文21篇，其中被SCI、EI收录论文8篇，申请软件著作权登记证5个。联合培养博士2人，其中1人已毕业；独立培养硕士生12人，其中6人已毕业。