Performance Evaluation of Preprocessing Techniques for Near-Infrared Spectroscopy Signals

Abstract Near-infrared spectroscopy is a non-destructive and firm analytical technique which provides unique information. Chemo metrics is a potential tool for both qualitative and quantitative analysis in the pharmaceutical industries, agro based chemical industries, petrochemical industry and in food industries. NIRS is a suitable method for analysis of solid, liquid and gas samples. Thus the Near Infrared spectroscopy is frequently used analytical method especially in food quality control laboratories. The food industry includes grading of fruits, seeds, spices, oil and milk. Adulteration is a serious issue faced by the consumers purchasing the food products. The adulteration aims at increasing the quantity of the food product and there by the quality is reduced automatically. Honey is one such product being contaminated by the sugar, beet and corn syrups in the market. The present work aims at deep learning approaches in studying the performance of the data obtained from NIR. Naive Gaussian, SVM, Logistic Regression, Gradient Boosting, Random Forest as classifiers and Accuracy, Precision, Recall, F1 score as parameters in performance evaluation in the present research work for the NIR signals of honey samples.