Search parameters

[Last updated 2018 10 06] When any of the four algorithm based searches are to be used then the first step is to some parameters about how the search will proceed. These are:

  1. Value to optimise: This is where you insert the model performance measure that needs to be maximised, by clicking on the value cell next to the chosen performance measure.
  2. Model size. This can be set in two ways:
    1. On the right of the Decision Tree search option is a drop-down menu allowing you to set the “depth” of the tree, meaning the maximum number of attributes in any o the models generated by the Decision Tree search
    2. Further below is the “Number of attributes in model configuration” setting, where you can specify the maximum number of attributes in any model found by exhaustive search
  3. Constraints: the project attributes and/or performance measures whose values must remain within specified limits. There are two ways of setting constraints:
    1. Click on one of the three options at the bottom of the Find New Models pop-up:
      1. Necessary and Sufficient. This enters FP=0 and FN=0 in the Constraints box.
      2. Necessary but not Sufficient. This enters FN=0 in the Constraints box.
      3. Sufficient but not Necessary. This enters FP=0 in the Constraints box.
    2. Enter any other preferred value in the Constraints box,
      1. These can refer to Confusion Matrix cell values e.g. FP<5,
      2. They can also refer to specific performance measure values, such as Lift>150%
    3. You can also go to the Design menu and in the list of Attributes Status select one or more attributes to have specific values (either 1 or 0)
      1. You can explore “the adjacent possible” by setting some of the attributes of a given model as a constraint and then exploring in the vicinity of that part of the model using “one additional attribute” search

Changing the Solver (evolutionary) search parameters

In the background, outside of EvalC3, there are settings which govern how Solver runs. To see and change these go up to the top of the Excel interface and click on Data, then look to the far right and click on Solver. Click on Options, then Evolutionary. For information on all the search parameters listed here, which you can change, see this web page:


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.