How to classifiy data using Fuzzy subtractive clustering?

Dear friends, Currently i am working in Fuzzy subtractive clustering. Belo i have mentioned the code. The problem i dont know how to get classification accuracy? I have created 5 rules using Fuzzy inference system.Can any one help how to implement Fuzzy rules using matlab code?? How to proceed after this?? Thanks in advance.

 finputtrain  = data(train,:);   % train_data
 foutputtrain = labels(train,:); % train_labels

 finputtest= data(test,:);        % test_data
 foutputtest= labels(test,:);     % Target

%clustering the data

[C,S] = subclust([finputtrain,foutputtrain],0.5);

% Generating FISuisng subtractive clustering

myfis = genfis2(finputtrain,foutputtrain,0.2,[],[1.25 0.5 0.15 0]);  

fuzout  = evalfis (finputtrain,myfis);
trnrmse = norm(fuzout-foutputtrain)/sqrt(length(fuzout));

testfuzout = evalfis (finputtest,myfis); 
testrmse   = norm(testfuzout-foutputtest)/sqrt(length(testfuzout));

Answer 1

[Prateekshya]

Hello Yuvaraj,

To calculate classification accuracy using a Fuzzy Inference System (FIS) generated from subtractive clustering, you should follow these steps:

• Evaluate the FIS on Test Data: You've already done this with evalfis, which generates fuzzy output predictions for your test data.
• Defuzzify the Output: If your FIS output is continuous, you might need to map it to discrete class labels. This often involves setting thresholds or using a method to determine which class a continuous output belongs to.
• Calculate Classification Accuracy: Compare the predicted class labels against the true labels to compute the accuracy.

Here is how you can implement these steps in MATLAB:

% Assuming 'testfuzout' is the fuzzy output for the test data
% and 'foutputtest' contains the true labels

% Step 2: Defuzzify the output
% For simplicity, let's assume you have two classes [0, 1]
% You might need to adjust this logic based on your specific classes
threshold = 0.5; % Example threshold for binary classification
predictedLabels = testfuzout >= threshold;

% Step 3: Calculate classification accuracy
correctPredictions = (predictedLabels == foutputtest);
accuracy = sum(correctPredictions) / length(foutputtest) * 100;

fprintf('Classification Accuracy: %.2f%%\n', accuracy);

I hope this helps!