|
Abstract:
|
A common situation faced by data professionals is that of receiving a large table of data, often containing many variables, and only a vague idea of what to do with it. For example, a large table of data containing information about insurance customers might arrive with a note to find out which customers are unprofitable. Even with a more focused query, it is useful to get a good look at the data initially before building a model. In this paper we will present a tool designed to give an initial look at data sets by using scagnostics to build an informal model of the relationships between variables. Based on the scagnostic analysis, a multipart visualization is built that is designed to show the main features of the data set. The Scagnostic AutoVis tool will be demonstrated, its design discussed and further enhancements suggested.
|
