histogram of inputs and targets
To get more insight into used data, it would be helpful to have information in the input and target distributions. The typical pairwise correlation diagram is not that meaningful from my perspective. But the distribution /histogram of each variable could be interesting.
create a plot for each station (not sure, because this will probably not be used at all, maybe only for some error checking if a specific station is much worse than others. But this wouldn't be an argument to spent always computation time and disk space for this plot.)CANCELED for now, but could be implemented if there is a request for it.
- create a plot for each input variable over all stations.
- create a plot for each target variable over all stations.
check out if this plot is can be created for transformed and untransformed data? (Then one could see that input data is adjusted to a Gaussian-like distribution)CANCELED for now, but could be implemented if there is a request for it.
- what about the filtered data? Is it possible to make this data visual too?
- what about different subsets (train, val, test)