Lectura de tesis
Tesis Doctoral
Lunes, 09 de Noviembre de 2026 10:00
- Título de la tesis: Exploratory Analysis of Collections of Process Variants from Event Logs.
- Doctoranda: María Salas Urbano.
- Directores: Cristina Cabanillas Macías y Manuel Resinas Arias de Reyna
- Tutor: Manuel Resinas Arias de Reyna
- Resumen: Process mining is a discipline that provides techniques for exploring, analyzing, and optimizing business processes from event logs. Event logs record all the information generated by information systems during the execution of the processes. By analyzing these event logs, process mining provides meaningful insights into process execution. These insights help organizations understand, analyze, and improve their business processes through tasks such as visualizing real process flows, identifying deviations from expected behavior, and detecting performance bottlenecks. Thus, process mining bridges the gap between data science and process management by providing objective, data-driven evidence of how processes are executed in practice. A key concept in this discipline is the process model, which represents the behavior of a process based on event data. Process models facilitate the understanding and analysis of business processes. Process mining techniques can automatically discover and enrich these models, which are often presented through advanced visualization methods that allow users to intuitively and interactively explore process behavior.
In real-world scenarios, the same process may be executed differently depending on the context, such as the organizational roles involved or the location where it is performed. This results in different process variants. Comparing the process models associated with these process variants is essential for identifying differences in behavior, performance, and outcomes, as well as for identifying patterns, deviations, and opportunities for improvement across different executions of the same process. However, current process mining techniques are not specifically designed to support the simultaneous analysis of multiple process variants and their corresponding process models. As a result, creating and comparing these models involves manual effort, making the analysis more time consuming and error-prone.
This dissertation aims to support process mining analysts by assisting them in managing collections of process variants and visually comparing multiple process models. To this end, we make the following contributions: i) we conduct a systematic analysis of process mining practices, identifying the central role of process models and comparative analysis in addressing specific business questions; ii) we identify fundamental tasks involved in analyzing process model collections and derive a set of functional requirements for supporting their creation, visualization, and exploration; iii) based on these requirements, we design an interactive system based on these requirements that supports analysts in creating and exploring process model collections, including mechanisms for identifying models that contain patterns of interest; iv) we create a process query language for creating and managing collections of process models according to some conditions defined by users; and v) finally, we identify key challenges in the visual comparison of multiple process models and propose a set of comparative strategies and visual designs to enhance the identification of relevant similarities and differences between models.
Overall, these contributions advance exploratory process mining by supporting the analysis of multiple process variants and their corresponding models. Rather than analyzing them individually, the proposed techniques enable analysts to work with them collectively, treating collections of process variants as first-class analytical objects for exploratory process mining.