7. Conclusion
The paper concludes that the fuzzy c-means clustering is an efficient dataminingmethod for selection of polymers for SAWsensor array coatings. In regard to chemical interactions with volatile organic analytes the method can efficiently segregate polymers with similar characteristics and provide the basis for polymer selection from dissimilar clusters. The fuzzy c-means clustering algorithm assumes that there are ‘c’ number of fuzzy clusters in the data and segregates data objects according to a defined fuzzy measure of similarity. The data objects in the present context are all polymers having selective chemical affinities toward target vapor constituents. A data point is representation of a polymer in terms of partition coefficients of all target vapor molecules. The partition coefficient matrix with polymers in rows and vapor molecules in columns defines the data. The polymer clustering is sought in data space defined by vapor partition coefficients as dimensions and polymers as data vectors. The subset of polymers representing the centers of ‘c’ fuzzy clusters is taken for the selection made for sensor array coatings. The optimum selection is attained by doing clustering analysis with successively increasing number of clusters until a common set of polymers emerges which make up for the
selection repeatedly for higher values of ‘c’. The sensor array simulation based validation ofthe present selection method by targeting detection of milk and fish freshness and spoilage may be adequately inspiring for the SAW electronic nose developers to provide low cost high performance solutions to food consumer safety. Acknowledgements The author Prabha Verma is thankful to the Council of Scientific and Industrial Research (CSIR, Government of India) for providing
Senior Research Fellowship for pursuing this work. The authors are thankful to all the contributors whose data were used for the analysis and validation in this work.