With the emergence of quantified-self and wearable computing, huge databases have been created to store an increasing flow of information. From these heterogeneous and ever expanding databases arise the need for new tools and methods that fit such data.
This three-years project is oriented toward the creation of a recommendation and exploration system. With instance-based algorithms suiting medical data, the aim is to provide an intuitive way for medical experts to visualize and understand their datasets. The system rely on automatic structuring and comparison of individuals to produce relevant answers to medical analyses.
By using instance-based learning, the originality of this work is to avoid overgeneralization and exclusion of atypical patients that are often overlooked by traditional approaches. All these features make the proposed tools and algorithms well suited for the medical field, where each patient is a special case. Leveraging the same reasoning formalized by the exemplar theory and used by professional experts, our prototype is intuitive for medical doctors and hospital staff.