| چکیده انگلیسی مقاله |
The E-nose technique usually comprises two main components, the first being the sensory unit and the next is the data processing unit which comprises machine learning algorithms to predict and classify the outcome. The oxygen-functionalized compounds are oxidized to acids on the surface of MoS2 QDs [1]. Here, combination of these catalytic properties of MoS2 QDs with the colorimetric responses of organic reagents was investigated to design a portable paper-based colorimetric sensor array for discrimination of cigarette smoke VOCs, as a complex matrix. First, the designed sensor array was used for classification of fourteen different cigarette VOCs. Principal component analysis (PCA) and linear discriminant analysis (LDA) for classification of data were used as unsupervised and supervised methods, respectively. In PCA analysis, the accuracy was 96 and 86% for the training set and cross-validation set, respectively. The LDA analysis was also performed on the raw data (accuracy was 100% in both training and cross-validation sets). The raw data obtained from image analysis were used as input in LDA, the quality of the row databases was good enough that was no need to pre-processing steps. In the second step, the sensor was validated by discrimination of five cigarette brands, the accuracy was found to be 100% for training set and 82 % for cross-validation set. In the third step, four of the best-selling brands in the Iranian market (Bahman Kootah, Omega, Montana gold, and Williams), all of which also had fraud samples (eight in total), were studied by the developed sensor array. The accuracy of discrimination of these four brands and their frauds were all equal to 100%. Next, all original and fraud samples (126 cigarette) were analyzed in a matrix with dimensions of (126 × 27) by LDA. The accuracy of this LDA model was 98%. Finally, the results of the measurements compared to the results originating by a standard analytical technique. FTIR spectroscopy of cigarette smoke was only able to discriminate between original and fraud cigarette samples, but did not provide an acceptable answer for classification of different brands of cigarettes. The analytical procedure proposed here, is fast, cheap, user-friendly, and reliable. The selectivity the developed sensor array could make it a valuable tool to differentiate original cigarettes from counterfeit products crossing every day national borders around the world. |
