Fast Electrochemical Impedance Measurement and Classification System Based on Machine Learning Algorithms

This work focuses on the analysis and detection of impedance changes in tomatoes using a wide-band Electrochemical Impedance Spectroscopy (EIS) system. Homogeneous tomato samples were subjected to EIS testing over a period of 72 hours to monitor the impedance variations during this time. The designed wide-band EIS system applied a structured signal to the tomato samples with a wide frequency range (1mHz to 1 MHz) and a maximum amplitude of 0.5V, capturing and storing the electrochemical impedance responses. Principal Component Analysis (PCA) was utilized to reduce the dimensionality of the dataset while retaining the most relevant features. Furthermore, three Machine Learning (ML) algorithms such as XGBoost, Artificial Neural Networks (ANN), and Support Vector Machines (SVM) were employed in the classification process. The process involved three key steps: data preprocessing, model training, and model testing. These algorithms produced a prediction model with a specialized feature selection that effectively identified the impedance changes in the EIS dataset. XGBoost and ANN demonstrated superior performance, successfully classifying all analyzed responses with 100% accuracy and effectively identifying the essential frequency points that significantly influenced the classification outcome using the XGBoost feature importance mechanism.

I. Introduction

Tomatoes, scientifically referred to as Solanum Lycopersicon, are a highly significant crop globally, particularly renowned for Italian production in both the European and international markets. The quality of fruits and vegetables has become increasingly important to consumers, leading to extensive efforts in monitoring their condition during storage, handling, and transportation [1]. Various techniques have been developed and employed to assess the health and quality of fruits and vegetables, including visual inspection, flotation separation, imaging sensors, and laboratory methods such as ethanol detection, mass spectroscopy, and fluorescence imaging [2]. However, many of these methods are complex, intrusive, costly, and time-consuming, typically requiring specialized laboratories and personnel.