Numeric boundaries for states

[gnopik]

Numeric boundaries are formed wrong for the states of input variables that have a limited number of unique values, which results in some missing and some 'overcrowded' scenarios:

Here is how the issue was solved in matlab:

[gnopik]

Some are already ok after the categorical fix, but some are still wrong: hype.csv - first two variables are outputs, no NaNs hype.csv

How the dashboard shows it

How it should be

[gnopik]

Another example where the dashboard defines states for the categorical variable wrong, misclassifying portions of data resulting in NaN scenarios (but no NaNs in the data). [converting [1 2 3 4 5] to [1000 2000 3000 4000 5000] did not help] data10.csv

[gnopik]

data.csv Another example, where the most influential categorical variable with five unique values gets misclassified into 4 states, the 5th remaining empty.