Confidence interval of ANFIS prediction

Is there a way to find the condifence values for predictions made by Neuro-Fuzzy designer?

[Sam Chak]

Can you show the mathematical formula for the Confidence Interval?

Answer 1

[Balavignesh]

Hi Akriti,

I understand that you want to find the 'confidence values' for a 'Neuro-Fuzzy Designer'.

The exact process of obtaining membership values may depend on the specific software you are using for Neuro-Fuzzy modelling.

However, a general approach to find 'confidence' values is to use the membership values associated with each prediction. 'Confidence values' can be calculated by normalizing the membership values. Membership values indicate the degree of 'confidence' in each prediction.

Here is an example code showing how you can do it:

prediction = 0.6; % Value predicted by Neuro-Fuzzy Designer
membership_values = [0.4,0.1,0.2]; % Example Membership Values

% Normalize Membership Values
confidence_values = membership_values/ sum(membership_values);

Please refer to the following documentation for more info on:

'Fuzzy Logic Designer' :

• https://in.mathworks.com/help/fuzzy/fuzzylogicdesigner-app.htm