Multiple files input to a Gaussian Progression Regression model in Regression learner app

I am working on a problem where I have a data set having 5 csv files in which each have 100 observations with 10 features and 1 response value. This data belong to device which fails after using it for some time, and 10 features are the time and frequency domain features extracted from a sensor installed on it. Now 5 csv files are for 5 similair devices, and data is structured in a format that response variable increases in a exponential way so a exponential trend is observed in it. All noise and pre-processing is done on it.

Now I would like to train a Gaussian Progression Regression model, on this data set. I have tried it on regression learner app and got some results, But my question is:

How should I create my dataset, should I concatenate all 5 files in one below otherlike stacking over other, but if I do so my validation and testing results are not so good. Other option could be to sort the data after stacking based on response variable values, which gives comparatively good results. But I am confused which way of stacking the data needs to be used to train the regression model ?