There is an optional Data Mining add on module for Decision Author which allows for the discovery of decision trees from historic data, and the possible deployment of these discovered trees as part of a runtime solution.
The above data mining tree shows the analysis of the propensity to accept / reject loan applications based on historic applications data. Overall 45.2% of the applications were accepted but the data mining analysis reveal 6 profiles with accept propensity ranging to 11.5% to 97%.
Data mining can be used to predict an event (credit default, occurrence of fault, lapsed insurance policies etc.) or it can be used to predict a numeric attribute (such as a process key performance indicator or measurement). This makes Data mining an important aspect of delivering operational intelligence solutions.
Decision trees like above can be generated automatically in one step or alternatively, the domain expert can be intelligently guided by the data mining algorithm to have the final say on the choice of attributes in the tree branching. This is done interactively by showing the user the importance of each attribute at a node given the information & statistical criteria used by the data mining algorithm (see below):