LAUSR

Abstract An eight metal oxide semiconductor sensor (MOS) based electronic nose (e-nose) has been used to characterize freshness of strawberry in different polymer package types. Pattern recognition methods such as principal component analysis (PCA), linear discriminant analysis (LDA), and support vector machine (SVM) were used to classify and investigate the effects of polymer packages on strawberry freshness. The packages were Ethylene Vinyl Alcohol (EVOH), Polypropylene (PPP), and Polyvinyl chloride (PVC). The response surface method (RSM) was considered for selection of optimized sensor array in terms of the contribution of each sensor in sample classification. Sample headspace patterns were examined on days 1, 8 and 16. The results revealed that PCA explains 84% of the variance between the data. The LDA classified samples with an accuracy of 86.4%. The SVM method with polynomial function could accurately recognize samples as C-SVM by 86.4% and 50.6% in training and validation, and as Nu-SVM by 85.2% and 55.6% in training and validation with a radial basis function, respectively. Finally, among the eight sensors used in the study, MQ8, MQ3, TGS813, MQ4, and MQ136 sensors were selected as optimal response sensors using RSM to reduce the cost of fabrication. Furthermore, optimal application sensors for each polymer package were identified using RSM.