Organizing, querying, and analyzing ad-hoc processes' data

Abstract

Business processes are central to the operation of both public and private organizations. Recently, business world is getting increasingly dynamic as various technologies such as social media and Web 2.0 have made dynamic processes more prevalent. For example, outsourcing and the emphasis on customer service makes the use of complex, dynamic and often knowledge intensive activities an inevitable task.

Ad-hoc processes, a special category of processes, have flexible underlying process definition where the control flow between activities cannot be modeled in advance but simply occurs during run time. In this dissertation, we investigate the problem of explorative querying, and analyzing of ad-hoc processes. Addressing this problem is challenging, as the information about process execution is scattered across several systems and data sources. Moreover, in many cases, there is no well-documented information on how this information is related to each other and to the overall business process of the enterprise. Enabling above-mentioned analysis requires a model and a query language for representing and querying process entities (e.g., events, artifacts, and actors), relationships among them, and the evolution of business artifacts over time. Moreover, the model should support multi-dimensional/-level views and analytics over ad-hoc processes data.

To address these challenges, we present a framework, simple abstractions and a language for the explorative querying and understanding of ad-hoc processes data from various user perspectives. We propose novel abstractions, folder and path, for facilitating the analysis of ad-hoc processes data by enabling process analysts to group related entities or find patterns among entities. We present FPSPARQL (Folder-, Path-enabled SPARQL) as a language and a set of new methods for organizing, indexing and querying ad-hoc processes. We then extend FPSPARQL for analyzing the evolution of process artifact, and for analyzing cross-cutting aspects in ad-hoc processes. We introduce two concepts of timed-folder to represent evolution of artifacts over time, and activity-path to represent the process which led to artifacts. Finally, we propose a model, GOLAP, and extend FPSPARQL for online analytical processing on process graphs. The approaches presented in this dissertation have been implemented in prototype tools, and experimentally validated on synthetic and real-world datasets.