4 Replies Latest reply on May 2, 2018 2:28 AM by Tanja Schneider

    Decision Tree with Text Nodes - improve looks and performance

    Tanja Schneider

      So, I built a Decision Tree for work, which startet out polygonic, but had to be downgraded due to performance issues. For this I relied on Olivier CATHERIN's post Decision trees, flow diagrams, sankeys in Tableau... here is a solution !!!, which was a huge help. I have a very large aggregated data base, and my decision tree has eight levels that can be chosen per parameter. Therefore, my tree can become extremely large if you're not applying a lot of filters. I chose to display all nodes as texts for better oversight. I switched the markers in the position sheets from "Gantt" to "Text", dragged the corresponding "level x" pill on "Text", and each node displayed its name directly.

       

      HOWEVER:

      Tableau seems to write the text of each node separately for every single leaf that's connected to this node, and position those texts on top of each other. As soon as a node has more than 2-3 leaves, the text begins to look bold and becomes blurry. Also, the QuickInfo only contains the volume information of the connected leaf at the very bottom instead of revealing the sum of volume for all its children. I fixed this by conducting a separate LoD on volume for each level, but this decreased my dashboards performance drastically. I attached a very similar workbook based on data and design of the workbook in the link above (but without my greedy LoDs to get the QuickInfo right).

       

      decisiontree_textnodes.PNG

       

      Does anyone know a way to build a decision tree with text nodes that only has one layer of text in each node? I would prefer a solution in which the volume information already contains the sum of all children, because I'd really like to get rid of those LoDs.

       

      Thanks in advance!

       

      Kind regards,

      Tanja