This thesis presented a visualization framework for designing process mining diagrams. The need of the framework was supported by a state of art research. The research showed a high importance of the visualizations in process mining field – most of process mining techniques use some form of visualizations. It also revealed the complexity of the designs. Every study that included a node-link diagram – the most common type of visualization –, presented it in a unique way. The third finding from the state of art research showed that regardless of the importance and complexity of the visualizations, most of the diagrams were designed without any help from visualization frameworks. Instead, the design decisions were based on a
combination of logical argumentation, existing practices and domain input. Some authors used also input from visualization theory, but did that in a fragmented way. Consequently, the research supported a need for a visualization framework for process mining diagrams.
The framework aims to identify the important aspects in interactive process maps that require conscious design decisions and provide information that helps the designer to justify their choices. It is based on two cornerstones – existing process mining visualizations and data visualization theory. Majority of the topics covered in the framework were extracted from a visualization theory by Tamara Munzner [37]. In addition, the idea of presenting the topics in a form of hierarchical questions stemmed from Munzner’s work [37]. However, adjustments were made to the base theory to make it relevant to process mining. Firstly, all the questions were retargeted to extract answers about the most common and complex form of charts in process mining – a node-link diagram, more specifically a process flow map. Questions were enriched with alternative answers that were relevant to process maps. The alternative answers were extracted from both, data visualization theory and existing process mining visualization practices. In addition, strengths and weaknesses of the alternative design choices were given where possible. These aspects were derived again from both, visualization theory as well as existing visualizations. Illustrations that are specific to process mining, were designed and added to the framework to increase the comprehensibility through visual examples. The framework was tested and evaluated in a case study. It was used by three participants, who were all involved in a project for developing a process mining tool for analyzing border crossing data. The participants used the framework for 25-45 minutes and shared their opinions in a semi-structured interview. The collected data was analyzed from the
perspectives of understandability, relevance, completeness and usefulness. Generally, the participants managed to use the framework without major struggles. The framework was found relevant and balanced in terms of how much effort it requires and how beneficial it is to the task at hand. However, understandability and completeness triggered several
suggestions. Some of the terms and targets of the questions were confusing for the
participants. Possible solutions were suggested, such as rewording the questions to clarify the targets and adding a glossary to define the terms. The main feedback on the completeness was about the missing aspects of the framework. The participants would have appreciated more alternative answers, more questions about designing embedded data and more
technique-specific questions, such as questions that could help to design deviances. The main value of the framework was found in making vague ideas concrete, coming up with new ideas and improving existing ones. The extended use was seen in developing the framework into a mock-up tool and using it for surveys to find out the users’ preferences for such diagrams.
The case study approach for validation was a good fit because the framework was tested out by the users and in the context the framework is developed for. However, the number of the participants was small and all the participants were part of the same project, which makes the generalizability of the feedback difficult. Another aspect that reduces the generalizability is the high variance of the feedback – most of the opinions were unique and depended on the individual who gave it. In further development of the framework, more feedback should be gathered from a higher number of participants in a variety of projects and some structured interview questions should be added to be able to make general conclusions out of the feedback. Nevertheless, the feedback that was gathered in the context of this thesis provides already valuable insights and ideas how to improve the framework on a cosmetic level as well as in general directions.
In addition to the ideas from the feedback, the framework can be developed further by expanding the defined scope of the framework. The framework can be extended to other types of charts besides process diagrams, such as various dashboards used in process mining tools. Also, it can be developed further to help with the full user interface design, not only the design of interactive process maps. Current version is tailored for static logs, which can be extended also to dynamic logs. In addition, the extension of the framework can be explored in terms of developing the diagrams for different types of screens besides laptops/desktops, for example taking in consideration characteristics of touchscreen design. The format of the framework can be improved by making it digital and taking advantage of the possibilities of interactive design, such as options to insert the answers interactively and produce automated reports of the collected data.