Authentication of Virgin Olive Oil by Using Dielectric spectroscopy combined with Some Artificial Intelligence Methods

Abstract

Adulteration is a serious problem in the food industry. Olive oil is widely adulterated with other cheap edible oils such as sunflower and canola oils. Therefore, developing a low-cost, practical and rapid analytical method for detecting such adulteration in olive oil would be useful and needed.  In this research, we aimed to develop a dielectric measurement based system combined with complementary analytical intelligent techniques to recognize authentication of virgin olive oil from adulterated with vegetable oils (canola and sunflower). 192 sinusoidal signals in the range of 20 kHz and 20 MHz were feed into the cylindrical dielectric sensor filled with oil sample. Correlation based feature selection (CFS) was applied to select the most appropriate dielectric features and eliminate irrelevant data. Support vector machines (SVMs), artificial neural networks (ANNs) and decision trees (DTs) were developed to classify virgin olive oil samples from adulterated ones. The obtained results indicated that ANN with topology of 2-5-3 has the best performance with accuracy of 100%.