Process Mining for Data-Aware Service Compositions

如何在开放、动态和不可控的环境下保证服务组合的可信度和质量是软件工程和服务计算领域的一个关键研究问题。作为构建服务组合并保证其可信性的重要手段，过程及其相应的过程挖掘技术受到了广泛的关注。然而，现有的流程挖掘方法大多侧重于控制流分析，无法满足与数据感知服务组合相关的技术需求。特别是在演化需求迫切的情况下，服务组合的事件日志往往不完整，严重影响了现有技术的挖掘质量。为了解决上述问题，本项目将软件分析与数据挖掘相结合，旨在研究基于事件日志中基本事件关系（即依赖、独立、互斥）的过程挖掘方法。更具体地说，我们希望在以下方面取得相当大的进展：最大化并发性和过程发现的块结构，基于跟踪依赖图的数据感知过程一致性检查，以及从事件日志中检测控制流错误。进一步，我们将在一些软件工具和相应的支撑平台上实现这些方法和技术，为提高服务组合的可信度和质量提供技术支持。

How to guarantee the trustworthiness and quality of service compositions in the open, dynamic, and uncontrollable environments is a key research issue in the domain of software engineering and services computing. As a signicant means to constructing service compositions and ensuring their trustworthiness, processes and the corresponding process mining techniques have received extensive attention. However, existing process mining approaches mostly focus on the control-flow analysis, failing to satisfy the technical requirements relevant to data-aware service compositions. Particularly, when the demand of evolution is urgent, the event logs of service compositions are usually incomplete, which signicantly affects the mining quality of existing techniques. To address the above problems, this project, by combining software analysis and data mining, aims to study process mining approaches based on essential event relations (i.e., dependences, independences, mutual-exclusions) in event logs. More specically, we hope to make considerable progress in the following aspects: maximizing concurrency and block-structuredness for process discovery, data-aware process conformance checking based on trace dependence graphs, as well as detection ofcontrol-flow errors from event logs. Furthermore, we will realize these approaches and techniques in some software tools and corresponding supporting platforms, in order to provide technical support for improving the trustworthiness and quality of service compositions.