For each model under consideration, it is a good idea to make a
methodical assessment based on the criteria generated in your test plan. Here is where you may add the generated model
to the stream and use evaluation charts or analysis nodes to analyze the effectiveness of the results. You should also
consider whether the results make logical sense or whether they are too simplistic for your business goals (for
example, a sequence that reveals purchases such as wine > wine > wine).
Once you’ve made an assessment,
rank the models in order based on both objective (model accuracy) and subjective (ease of use or interpretation of
results) criteria.
If you were able to address the above issues successfully and believe that the current models meet your goals, it’s
time to move on to a more thorough evaluation of the models and a final deployment. Otherwise, take what you’ve learned
and rerun the models with adjusted parameter settings.
|
