Prediction on the Level of Toxicity in Fruits and Vegetables based on PAHs using Machine Learning

Abstract

This study focuses on assessing the toxicity levels in fruits and vegetables based on the presence of polycyclic aromatic hydrocarbons (PAHs), particularly in regions affected by industrial and vehicular pollution. The presence of PAHs, primarily due to air pollution particulate matter deposition on plant surfaces, is a critical concern for the nutritional quality of fruits and vegetables. Traditional methods, including Gas Chromatography/Mass Spectrometry (GC/MS) and High-Performance Liquid Chromatography (HPLC), are conventionally employed to measure PAH levels in fruits and vegetables. These methods are found to be valuable but expensive and time-consuming. Although, the detection of toxicity relies on either expert knowledge or experimental analysis when compared with the limitations set by EFSA (European Food Safety Authority). Therefore, in this study, artificial intelligence techniques have been employed to evaluate the toxicity levels based on 16 PAHs. The PAHs concentrations in fruits and vegetables were collected from different articles corresponding to safe and unsafe dataset, then validated through statistical analysis. The validated dataset is classified using different machine learning algorithms. Based on the output from neural network, the level of toxicity is also scaled and compared with the targeted outputs. The promising results of the classification of toxicity using artificial intelligence methods are substantiated by an experimental study and validated through statistical methods. From the results, it can be observed that the machine learning algorithm has given classification accuracy more than 90% along with their degree of harmfulness. This research holds implications for food safety and public health, offering a novel approach to the interdisciplinary understanding of climate change by addressing the impact of environmental contaminants on the edibility of fruits and vegetables.