How do i know what TF model to use?

General Discussion
, ,
#1

Hello. I am very, very (1 week) new to ML and TF.
I have a dataset of many measurements from different production-steps of different times of a company.
Some of these measurements have to be in a certain intervall to be labled “ok”. Othervise they indicate that this production step is broken.
I want to take these measurements (from an interval of for example 1 year) and predict when in the future they will deviate so much, that they are out of their interval.

I am doing an internship and i am a little bit stressed. In the internet i can only find tutorials as in “how to build this TF model” but never how to even know what model to use…
The only thing i could think of here was Regression? But i really don´t know.

If anyone could give me a hint, that would be awesome!

(I am not a nativ english speaker, so sorry for anything wrong with the sentences)

#3

We have some time-series tutorials/examples at:

1 Like
#4

It could be a regression. It could be a classification problem. Depends on how you structure your data and labels.
Using a regression model you can predict how many days into the future the system will break.
But you can use binary classification models as well: given the previous condition of the system (previous data), how likely it is that the system will break within the next week or a month (model outputs 0 or 1 for each future period).