While learning logistic regression, I had question that while evaluating model only considering accuracy or error rate is fine or should we have any other parameter to consider. I an got to know that, we should consider Kappa value to check the robustness of the model means accuracy we are getting is not by chance.
What is Kappa Coefficient or Cohen's Kappa value?
It is used to calculate the agreement between the two types of values in our case actual values and predicted values. It compares observed/predicted accuracy with the expected/actual accu
Kappa = (Observed accuracy - Expected accuracy)/(1 - Expected accuracy)
Let us consider following contingency table where we have actual and predicted values.
Poor agreement = Less than 0.20
Fair agreement = 0.20 to 0.40
Moderate agreement = 0.40 to 0.60
Good agreement = 0.60 to 0.80
Very good agreement = 0.80 to 1.00