SlidingWindow average model?
Liliput
New Altair Community Member
OK here is my question:
In a validation operator like the "Sliding Window" validation i get the performance averaged for all predictions during the slide. This is fine but is there a way to average somehow the rules the model uses to predict? I mean i had placed the "RuleLearner" to work and a "ResultWrier" after it. So i can see the genarated rules for every iteration fro the validation but i want an operator to average the best mix of all generated posibilities. Something like learn from all interations to be best.
Which is i should use, please?
Txh
In a validation operator like the "Sliding Window" validation i get the performance averaged for all predictions during the slide. This is fine but is there a way to average somehow the rules the model uses to predict? I mean i had placed the "RuleLearner" to work and a "ResultWrier" after it. So i can see the genarated rules for every iteration fro the validation but i want an operator to average the best mix of all generated posibilities. Something like learn from all interations to be best.
Which is i should use, please?Txh
Tagged:
0
Answers
-
Hi,
why don't you transform your time series to an exampleSet containing all the windows. Then you could use the complete series for learning. The learner will probably handle the additional information better than simply "averaging" the resulting rules.
But if you prefere to do it this way: Use the bagging Operator and load the models within its child operators. The result should give you a combined model averaging the confidences delivered by every single model.
Greetings,
Sebastian0
