The effect of data fusion on improving the accuracy of olive oil quality measurement

Abstract

Olive oil is one of the healthiest and most nutritious edible oils, and it has a great potential to be adulterated. In this research, fraud samples of olive oil were detected with six different classification models by fusion of two methods of E-nose and ultrasound. The samples were prepared in six categories of adulteration. The E-nose system included eight various sensors. 2 MHz probes were used in through transmission ultrasound system. Principal Component Analysis method was used to reduce features and six classification models were used for classification. Feature with the greatest influence in the classification was “percentage of ultrasonic amplitude loss.” It was found that the ultrasound system's data had worked more effectively than the E-nose system. Results showed that the ANN method was recognized as the most effective classifier with the highest accuracy (95.51%). The accuracy of classification in all the classification models significantly increased with data fusion.