Can Knime do Scenario Decision Analysis?

I’m looking for some sample workflows or suggested approaches to evaluate resources, constraints and objectives as inputs and output a set of ranked scenarios by running the permutations. I’m thinking decision tree with random forest as an initial approach. I only have inputs though and not historical data to train and validate as the outputs are all future scenarios. The net objective is to pick the scenario with the most likely favorable result. Any suggestions?

Might the Parameter Optimization nodes work (loop start and end)? You will need to define what you are optimizing for. This would help you define the set of parameters to optimize a given objective.

1 Like

I only know Decision Tree usecases in KNIME for ML predictions not the way I understand your description. However if you manage to frame your problem in an optimization problem evaluating a numerical target with RMSE MAE or other error metric (basically some kind of regression) then you could use @supersharp 's proposal and use your features as inputs for the parameter loop and see what gives you most ■■■■ for your buck

Thanks Supersharp for buying me a vowel. I did some investigation into the optimization extension and created a very basic workflow that demonstrates the concept I was looking for. I think it will work if I can properly define the inputs, constraints and desired outcome. In my simple test, I was able to iterate through the permutations, produce a list of results that satisfied the test requirement and identify the best result from those.

1 Like

Great to hear! For the constraints, you could probably put in a rule engine and some try-catch nodes (if inside the loop). In the cases where the ‘rule’ (i.e. constraint) fails, the catch option will send through an alternative result (e.g. 0) to continue the loop. Or, simpler, but maybe more compute heavy - run all permutations and after the loop drop off the cases which do not meet the constraints with a rule-based row filter.

Feel free to share with the community on the hub if you get something good going! Good luck!

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.