Visually Mining on Multiple Relational Tables at Once

Data mining (DM) processes require data to be supplied in only one table or data file. Therefore, data stored in multiple relations of relational databases must be joined before submission to DM analysis. A problem faced during this preparation step is that, most of the times, the analyst does not have a clear idea of what portions of data should be mined. This paper reckons the strong human ability to interpret data in graphical format to develop a process called “wagging”, to visualize data from multiple relations, helping the analyst when preparing data to DM. The data obtained from the wagging process allow to execute further processes as if they were operating over multiple relations, bringing the join operations to become part of the data mining process. Experimental evaluation shows that the wagging process reduces the join cost significantly, turning it possible to visually explore data from multiple tables interactively.