Freshness Analysis of Orange using Bio-impedance Measurement and Machine Learning Algorithms

Assessing the quality of fruits after harvesting has become crucial to satisfy both production and consumer requirements. The freshness of fruits is measured with the help of various non-destructive methods to provide quality attributes without damaging their quality and texture. The utilization of computational techniques, such as Machine Learning (ML), is considered suitable for identifying physiological and bio/chemical modifications in fruit. This study employed the Electrochemical Impedance Spectroscopy (EIS) technique to evaluate the freshness of the orange. The bio-impedance dataset was obtained for six days using the AD5933-based measurement setup. Various classification and regression-based ML algorithms were employed using MATLAB to predict the freshness of the orange. The ML classifiers and regressors were further evaluated based on their various performance indicators. In classification, the Support Vector Machine (SVM) and Neural Network (NN) demonstrated better performance for prediction. Whereas the NN regressor and Gaussian Process Regressor (GPR) algorithms exhibited better performance in precisely predicting the freshness of orange.