How is the ground truth 1 or 0 being created in the train/test set,does an expert create that or is it created by an existing algorithm. If that is already an existing algorithm,then model which I build should be able to capture those structural breaks which the existing algorithm has not been able to capture,how is that going to be validated?
All of the information we have been provided by the client are on the overview page.
To score the models during the out of sample phase, we kept another 10’000 datasets that your code will be run against only once.