Using $$\varphi -\delta $$ Diagrams on Web Data

As the Web is still expanding, also the demand for fast and accurate tools aimed at analyzing digital documents (e.g., webpages) is constantly growing. In this scenario, the main strategies for producing accurate predictive models are mostly focused on the assessment of classifiers performances and features. In this work, a graphical tool, called \( \varphi - \delta \) diagrams, is applied to some use cases aimed at highlighting its potential in supporting development and implementation of Web systems and services. In particular, \( \varphi - \delta \) diagrams permit to visualize (i) classifier performance, in terms of accuracy and bias, and (ii) variable importance, useful to define feature ranking, selection or reduction algorithms. The proposed use cases emphasize the usefulness of the tool when dealing with Web data.