Fuzzy c-means clustering based outlier detection for SAW electronic nose

The paper presents a novel way of using fuzzy c means clustering in supervised manner for outlier detection in surface acoustic wave sensor array based electronic nose. The method is trained through the set of known class labels samples taken as train set combined with the test set samples, which may contain outlier class and/or true data samples. Fuzzy c means clustering is implemented with equating cluster number equal to one class greater than the true data classes actually present in the train data set. A criteria of outlier detection based on the sum of membership grades of the train data set samples has been developed. The obtained results show that the proposed fuzzy c means clustering based outlier detection method detects the true data samples and outliers quite efficiently.