How to build a model when the training sample consists of every minute readings?

Question

Python

I don’t know how to approach the issue of building a model: the
training sample consists of every minute readings (process sensors that last an hour), and the target variable (quality) is issued every hour, respectively. That is, for each target variable there are 60 rows identical in the data set.

Answer 1

training set = { ([ process parameters over time], quality) }

Answer 2

In your task, most likely, the quality indicator somehow depends on changes in the process parameters during the previous hour (or maybe - not only one, but also several previous hours - the so-called systems with a delay. Or maybe - and even more intricate - dependence there is also the value of the quality indicator in some previous periods). For example - were there any emissions, were there cyclical changes, etc. There are many such process indicators. From trivial - mean, standard deviation, to more complex - the form of the distribution function of values, entropy, the presence of a trend and its characteristics, etc. And most likely from some of these indicators and / or their combinations, as well as from mutual changes in the indicators of the time series you observe, somehow depends on the value of the quality indicator, which you receive every hour. This is exactly what you need to find out. Those. analyze the process per hour (this is your input) - make a quality prediction (this is the output of your model). Well, you can, of course, stupidly try to stuff all your raw data into some kind of grid (N * 60 points per hour, where N is the number of your sensors), but I think this will be of little use. Therefore, obviously, the first stage that you will have to go through is Feature engineering, i.e. constructing parameters, according to which you will then train your model. where N is the number of your sensors), but I think this will be of little use. Therefore, obviously, the first stage that you will have to go through is Feature engineering, i.e. constructing parameters, according to which you will then train your model. where N is the number of your sensors), but I think this will be of little use. Therefore, obviously, the first stage that you will have to go through is Feature engineering, i.e. constructing parameters, according to which you will then train your model.
Something like that.