Visual data mining tools integrate machine learning, information visualization, and human pattern recognition capabilities for effective data exploration and analysis. When data is large and high dimensional, humans cannot easily derive conclusions from it, which means data must often be reduced and abstracted. Generally, domain experts are not involved in data abstraction steps and are not informed of the resulting information loss, which can affect informed decision-making.
To evaluate how useful it is to involve domain experts and their knowledge in data abstraction steps, which are required to understand and visualize information, we conducted a case study with the Calgary Epilepsy Program. We designed CEP-Vis, a patient centric visual analytics tool, using a user-centered approach to allow clinicians to compare a patient to other patients with epilepsy. We evaluated CEP-Vis through usability studies and interviews and presented the results and challenges of integrating this tool in a real work setting.