Options
Error in Binomial Performance Classification Operator
Dears
I use deep learning for time series prediction in a binary classification problem. The result sometimes just contains a value for prediction (e.g., just "false"). The binominal classification performance operator has an error in this scenario with this message:
"The attribute prediction has 1 different values, must be 2 for calculation of precision."
I removed the precision metric, but it has the same error for false_positive metric:
"The attribute prediction has 1 different values, must be 2 for calculation of false_positive."
This is unacceptable that this operator can't calculate this metric! It should consider 0 for the value. Besides, I can't find any operator to add the "true" value to the annotation of the prediction attribute. Is there any operator or trick for this problem?
Sincerely
I use deep learning for time series prediction in a binary classification problem. The result sometimes just contains a value for prediction (e.g., just "false"). The binominal classification performance operator has an error in this scenario with this message:
"The attribute prediction has 1 different values, must be 2 for calculation of precision."
I removed the precision metric, but it has the same error for false_positive metric:
"The attribute prediction has 1 different values, must be 2 for calculation of false_positive."
This is unacceptable that this operator can't calculate this metric! It should consider 0 for the value. Besides, I can't find any operator to add the "true" value to the annotation of the prediction attribute. Is there any operator or trick for this problem?
Sincerely
Tagged:
0
Best Answer

OptionsSabaRG Member Posts: 13 Contributor IISorry Dears
I already checked many solutions like using replace operator, but finally, I found a solution to use "set positive value" operator and define the "true" value as the positive value in which the performance operator can calculate the false_positive, true_positive, precision, f_measure and other values.
Anyway, I believe the performance operator must caluclate the false_positive and true_positive values as zero and log a warning for this condition or has an option for this case which users can select ignoring option or error.
Sincerely0