Decision Tree: How do I change the default Options?

Default settings for Decision Trees can be set from the main Options control. You access this via the main Tools...Options menu and selecting the Decision Tree tab:

Modifying these settings will only affect subsequent Decision Trees and will not affect any that have already been built.  These settings are saved and will be retained by subsequent sessions.  These settings will be the initial settings for all new Decision Trees.  The settings can be altered for individual Decision Trees once they are created.

The initial Marketing Costs and Stopping Conditions can be set using the main Options panel.   

Default colours used in the different displays can also be set here:

  • Analysis Colour used in Organic Tree, Box Tree, Next Splits and torus

  • Non-Analysis Colour used in Organic Tree, Box Tree, Next Splits and torus

  • Neutral Colour used in Organic Tree, Box Tree and Next Splits

  • Background Colour used in Organic Tree, Box tree and Next Splits

  • Focus Node Highlight used in Tree Grid, Organic Tree and Box Tree

The "Reset Decision Tree Options" button restores all the settings to the original defaults.  It also resets other properties which can be configured in the interface.  For example:

  • Column Layout in the grid displays

  • Active tabs

  • Button settings such as Simplify Query, Tree Grid layout

System Memory and Batch Size

Performance issues can arise while building a Decision Tree when multiple nodes are processed at the same time.  This occurs when you use the "All Leaves" algorithm option, and when dimensions are added to a partially built tree as all leaf nodes have to be updated.

Memory issues are only likely to be a problem if you have a very large database (e.g. 10 million records or more) or are using high cardinality variables (e.g. Postal Sector with 10,000 categories).

The batch size can be limited in 2 ways, and whichever is the most restrictive is applied:

  • The number of nodes processed together is limited so that memory usage stays within the limit set by the "Maximum batch memory (in Mb)".  

  • You can artificially limit the number of nodes using the "Maximum batch size" option.  

The batch size used in processing is visible in the status bar when the tree is building.  In this example, the batch size is 8:

Further Changes

Further changes can be made to the settings of individual trees by using the properties window accessed via "Views...Properties".  This is a less user-friendly interface but provides greater flexibility.  Settings made here only affect the current Decision Tree.