Welcome to the BaVA research group!
Over the last two decades, simple concepts in Exploratory Data Analysis (EDA) (Tukey, 1977) have extended to complex methods in data analytics (DA). The purpose of DA is to uncover known and/or unknown data structures, such as clusters, trends, and relation-based networks, that domain experts may assess. The hope is that experts learn, i.e., update, adjust, or confirm their current scientific judgements, from the data based on their assessments of the DA summaries. However, most DA strategies operate independently of the learning process and lack transparency. Hence, we introduce BaVA. BaVA offers a framework to transform standard DA methods into interactive data exploration approaches that inherently accounts for the expert.
BaVA combines Visual Analytics (Card et al., 1999; Pirolli and Card, 2005; Thomas and Cook, 2005) and Bayesian statistics (Bernardo and Smith, 1994) to transform standard analytic methods into interactive data exploration approaches.
Visual to Parametric Interaction (V2PI) is a non-probabilistic version of BaVA. Please explore this website to learn more about BaVA and V2PI.