VizRec is a service that learns from user choices to suggest the best possible visualization for any datasets, simplifying interactive analytic tasks.
The visualizations that VizRec suggests are perceptually correct, because it uses a rule based system matching data attributes with visual variables of visualizations to generate them (Mutlu et al. 2014KC). Furthermore, VizRec accounts for the variability in preferences for choosing even simple charts (Mutlu, Trattner, Veas 2015KC). It leverages a recommender stage built upon a large crowd-sourced study, that continues to learn user preferences (Mutlu et al., 2016KC). Using information theoretic measures, VizRec ensures that: i) the appropriate sources of information are used to describe visualizations, and ii) the methods produce the most suitable recommendations thereby (Mutlu et al., 2017KC).