Greetings I have a question, about the penultimate function in the code, the result it generates is something random that out of 1.2 million predicted rows takes 15 randomly, in this way it cannot be predicted when, it will happen because the target event may be predicted correctly but with this function that takes randomly it will not be taken into account so my question is more important the date or the label of the event or in what proportion the evaluation is important
Posted by: carlosv0410 @ Sept. 23, 2021, 4:47 a.m.Yes, sure it does.
However, purpose of that notebook is to show how to do the submission rather than how to define the workflow for your prediction, consider that you have lots of possibilities to build your own model, maybe another type of classifier or some architecture of NN. Currently, we have provided a workflow you may use to structure your datasets, thus you will have more time to do some research on the literature for overcoming this challenge.
Posted by: bluemirrors @ Sept. 23, 2021, 8:03 p.m.By the way, event date estimation is always the most important issue for a predictive maintenance task.
Cheers.