Options
2024
Doctoral Thesis
Title
New Data Sources for Process Mining
Abstract
Process mining is situated between process science and data science, with the objective of discovering, monitoring, and improving processes. As a technique in business process management, process mining uses event logs as input and enables building a view of reality by making process executions tangible, facilitating the identification of bottlenecks, providing new insights, and anticipating problems in business processes. The process mining discipline has undergone a notable evolution over recent decades, both in academic research and in practice. The development of innovative algorithms and algorithmic extensions is an ongoing process, with previously unused data sources being explored to enhance the existing body of knowledge. Traditionally, the predominant data sources for process mining are information systems, such as enterprise resource planning systems. However, process information can also be extracted from alternative sources and utilized for process mining. The majority of available data is unstructured, thereby offering a substantial opportunity to provide valuable contextual insights into business processes. This can facilitate a more comprehensive representation and analysis of real-world processes, as well as the reduction of blind spots, i.e., parts of processes that were previously unable to be captured in event logs. The overarching objective of this dissertation is to contribute to the advancement of process mining by enabling the use of new data sources. This objective is pursued by facilitating the integration of sensor, video, bot, and text data, as well as providing a systematic overview of approaches to unstructured data in process mining. In line with design science research principles, multiple artifacts are developed that contribute to process mining in both research and practice. This dissertation comprises five research articles addressing three opportunities that aim to expand the scope of process mining analysis. First, the integration of unstructured data into process mining is addressed by initial research. However, a systematic overview that presents a comprehensive summary of the approaches employed has yet to be provided. Research Article 1 addresses this opportunity by providing a systematic literature review of the current state of research on the use of unstructured data in process mining. In light of the findings, a research agenda is put forth that identifies both open challenges and potential avenues for future research. Second, data derived from video and sensor data could facilitate the detection of previously hidden but relevant process activities, thus enabling a more transparent process picture. Accordingly, Research Article 2 presents a reference architecture that offers guidance on the utilization of unstructured data sources and traditional event logs for object-centric process mining. Moreover, an instantiation of the proposed architecture is provided, demonstrating the specific use of video and sensor data for object-centric process mining. Additionally, Research Article 3 proposes a reference architecture for the unsupervised exploration of video data. This architecture enables the extraction of actual process activities from video data without the need for predefined activities, serving as a starting point for process discovery. Third, the conjunction of several emerging technologies with process mining has facilitated the integration of additional data sources. As the use of Robotic Process Automation (RPA) bots becomes more prevalent in business processes, there is a need to integrate the steps performed by bots into process mining analysis. Accordingly, Research Article 4 presents an approach that makes bot logs from RPA software usable for process mining and develops process mining measures that analyze bot logs and process event logs in an integrated manner. Chatbots represent a further type of bot that is deployed in scenarios where it is essential to align with the underlying business processes and comply with regulatory requirements. Consequently, there is a need to integrate textual conversation data from chatbots to investigate whether chatbots achieve process-compliant behavior. Research Article 5 addresses this opportunity by providing an approach that converts textual conversation data from chatbots to event logs for process mining and quantifies chatbots’ ability to learn and adhere to organizations’ business processes. Overall, this dissertation, comprising the five research articles, contributes to the advancement of process mining by enabling the use of new data sources. The conducted literature review provides new insights to systematically advance research at the intersection of unstructured data and process mining. The dissertation presents two artifacts that provide guidance for incorporating video and sensor data as new data sources for process mining, completed by a generic architecture on the use of unstructured data for object-centric process mining. Finally, two artifacts that integrate data from emerging technologies are presented. Both artifacts facilitate a more holistic process view by incorporating data from chatbots and RPA bots as new data sources for process mining.
Thesis Note
Bayreuth, Univ., Diss., 2024