analyte-basicity/polymer-acidity interaction, and l and log L16 represent combined cavity creation and other dispersion effects. In the LSER model, individual analytes and polymers are assigned independent sets of values for their solvation parameters. That is, the partition coefficient for any analyte-polymer combination can be found if their solvation parameters are known. In Tables 1–3 are listed the values of these solvation parameters for various milk and fish marker VOCs and for the prospective sensing polymers. This data has been collected from various published sources as mentioned there. Based on these tables one can construct the K-matrix with the vapors in rows and the polymers in columns. Each row in the K-matrix then specifies a particular vapor by its partition coefficient values in all polymers, or equivalently by the responses of sensors in an array prepared by these polymers as sensors coating. The listed set of prospective polymers would however be invariably large, and the sensor array responses may contain lots of overlapping information about vapors. The task of polymer selection is to select smallest number of polymers that can generate maximally discriminant information about the target vapors. In Section 4 we present a fuzzy c-means clustering based
method for polymer selection. This is followed by the validation of this selection in Section 5 based on the SAW sensor array response simulation and detection analysis for the milk and fish freshness and spoilage. Tables 4(A, B) and Table 5 present the partition coefficient data for the milk and fish VOCs markers respectively based on the solvation parameters data in Tables 1–3 and the LSER equation (1). The data are presented in the form of transpose of K-matrices (denoted as KT-matrices: polymer in rows and vapors in columns). This data will be used in following section for polymer selection.