|
Abstract:
|
We wish to read datasets and display interesting aspects of their content. The design to do this rests on the grammar of graphics, scagnostics, and a modeler based on the logic of statistical analysis. We distinguish an automatic visualization system (AVS) from an automated visualization system. The former automatically makes decisions about what is to be visualized. The latter is a programming system for automating the production of charts, graphs and visualizations. An AVS is designed to provide a first glance at data before modeling and analysis are done. AVS is designed to protect researchers from ignoring missing data, outliers, miscodes and other anomalies that can violate statistical assumptions or otherwise jeopardize the validity of models. This capability is especially important for machine learning problems on big data, where examining all possible variables (in the thousands, say) and cases (in the billions, say) is practically impossible.
|
