Comparison Analysis of Machine Learning Models for Identification of Pesticide Residue

Pesticides are chemical substances that are regularly used by farmers to enhance crop yields through controlling pests that harm plants. However, despite the crucial role that pesticides play in agriculture, they pose potential risks to the environment and may have chronic effects on human health and hence raise key concerns about food safety. Therefore, adopting food control systems to identify pesticide residue in food is of major importance. This study explored the capabilities of various machine learning models to detect pesticide residue in food samples. To this end, the study considered data collected from Bahrain’s Pesticide Residue Monitoring Program spanning the period 2015–2019. For comparison purposes, different classification methods have been implemented covering K-Nearest Neighbors, Logistic Regression, Quadratic Discriminant Analysis, Naïve Bayes, and Support Vector Machine. The conducted analysis revealed that machine learning models can effectively detect pesticide residue with an overall accuracy of 75%. However, the performance of the adopted models can be improved by different directions such as introducing relevant variables, addressing imbalanced data issues, and considering advanced deep learning models.