Discussion:
[Scikit-learn-general] How to feed event output back into classifier?
Startup Hire
2016-02-15 09:50:20 UTC
Permalink
Hi all,

Hope you are doing good.

I am using a OneVsRestClassifier with SGD as estimator to rank-order of 5
different lables using the predict_proba methods.

The ranked labels are actually products. The rank wiill be used to decide
the order in which the customer view the products.

My question is this:

Is there a way by which I can include my conversion event back into my
model?

Say, my rank order was A,C,B,.. which means I expect customer to click and
probably buy A.

Probably they end up buying C in most of teh cases. This kind of feedback
should be iteratively fed into the model so that it adjust accordingly.

(or)

Do you think if I constantly re-train the model at regular intervals taking
in recent data, this would be taken care?

Thanks,
Sanant

Loading...