Predicting antibacterial effects of ozone treatment on green leafy vegetables using machine learning models

Abstract This study investigated the antibacterial and color effects of ozone treatment on green leafy vegetables, a crucial factor for improving food safety and quality. Machine Learning (ML) models were employed to accurately predict the outcomes of this treatment, offering a more efficient alternative to traditional methods. The dataset, comprising 720 data points, was generated using ozone concentrations (2, 5, and 10 mg/L and treatment times (0, 5, 10, and 15 min) to measure microbial log reduction. Ozone concentration significantly affected bacterial reduction, while vegetable varieties influenced color parameters measured in the CIELAB color space. However, ozone concentration and treatment duration had no effect on color. Five ML algorithms were tested to evaluate prediction accuracy, with Random Forest (RF) emerging as the best regressor. RF achieved a test accuracy score of 0.96 after optimizing parameters through 5-fold cross-validation. Feature importance analysis identified ozone concentration as the most critical variable for predicting outcomes. The findings indicate that ozonated water effectively decontaminates leafy greens, maintaining microbiological quality and color. Additionally, ML methods can successfully model the key features of this treatment.