Ripening Assessment Classification using Artificial Intelligence Algorithms with Electrochemical Impedance Spectroscopy Data

Abstract

The increasingly growth of the world’s population requires more efficient efforts to manage people’s needs. One important problem lies on food distribution, quality and sustainability throughout the chain, from the producer to the consumer’s hands. It’s more and more important to achieve intelligent non-destructive methods to assess food quality, at the crop, storage and shelves at the supermarket. This paper approaches an application of machine learning (ML) on impedance data from bananas to assess whether the fruit is good for consumption or not. To fulfill the proposed objective, impedance data of 10 bananas during 32 days was acquired through electrochemical impedance spectroscopy (EIS), using an Analog Discovery 2. A database was produced with impedance, humidity and temperature values from each measurement. After data pre-processing, several ML classifiers were trained and tested for several different feature combinations and data normalization methods. The XGB classifier achieved the best performance, with a F1-score of 98,36% and accuracy of 98,10%.This study can be extrapolated to other fruits and vegetables in order to allow a better management in the food industry, improve quality and prevent food waste.