| dc.description.abstract | Context - Process Mining is a discipline that intersects business process management and data science. Event logs are extracted from information systems and analyzed through various techniques in data science to optimize business processes and improve user experience or process efficiency. Data preprocessing is an important part of data analysis, which often occupies most of the time in the analysis process.
Purpose - In most of the existing studies, we could see the descriptions of the event log preprocessing techniques used in specific case studies. However, only one research classified event log preprocessing techniques, but this taxonomy is not task-oriented and operation-based and does not include all preprocessing techniques. So this thesis project aims to propose a new taxonomy while exploring the relationship between the choice of data preprocessing techniques and domain, data domain, analysis purpose, and process mining task.
Method and result - A two-phase research method, systematic literature review (SLR), and taxonomy development are applied in this thesis project. A taxonomy of event log preprocessing techniques with six high-level categories
and twenty low-level categories is proposed. The high-level category consists of log filtering, log transformation, log abstraction, log integration, log enriching, and log reduction. The results of SLR show that the selection of log filtering is not data domain-dependent, whereas the choice of other log preprocessing techniques is. Data domain and analysis purpose affect the log preprocessing techniques selection, and the intended PM tasks do not have a significant impact on it.
Contribution - For scientific contribution, this research will fill in the gaps in academic research and update existing basic operations of event log preprocessing; For practical contribution, this research will provide insights to analysts in making preprocessing technique selection decisions in different task contexts. | |
