Identification of artificially ripened banana by chemical agent using gas sensor array

Fresh fruits are consumed for the necessary vitamins, minerals, fiber, and proteins to keep ourselves healthy. The banana is a common and healthy fruit produced around the year in Bangladesh and many countries in the world. Bananas are often artificially ripened by chemical agents, which is injurious to health. Therefore, detecting artificially ripened bananas (ARB) has become important in ensuring food safety and quality. A novel non-destructive approach to identify ARB by analyzing electronic signals acquired by a gas sensor array in response to volatile organic compounds (VOCs) emitted from bananas is proposed in this paper. An automated data collection system is designed, which consists of a gas sensor array, a data acquisition device, and LabVIEW. Four machine learning (ML) models, namely logistic regression (LR), support vector machine (SVM), decision tree (DT), and Extreme Gradient Boosting (XGBoost), have been utilized to analyze the sensor data and classify the bananas into both binary and day-wise (ripening stages) classification of artificially and naturally ripened banana (ANRB) tasks. Among the four models, XGBoost achieved the best performance on binary classification of ANRB with 97.03% and 95.05 ± 2.98% test and 10-fold cross-validation (CV) accuracy while incorporating important features. As for the day-wise classification, the SVM model performed best, obtaining 99.94% test accuracy and 99.92 ± 0.05% CV accuracy (utilizing all features). Moreover, this study is the first investigation of ANRB detection using VOC response, which also outperforms existing hyperspectral imaging-based ANRB classification in a substantially cost-effective manner. Overall, this study contributes to the advancement of food safety technologies by providing a non-destructive technique, an alternative to destructive solutions, for ARB detection.