Safety risk assessment and early warning of chemical contamination in vegetable oil.

Edible vegetable oil is a necessity in people's daily diet, and its safety has attracted the attention of the government and consumers. In recent years, the safety incidents of edible vegetable oil are reported, and the safety problems of edible vegetable oil caused by chemical hazards such as benzopyrene, heavy metal and aflatoxin B1 are more serious. In this research, a risk assessment model through dietary exposure assessment and margin of exposure (MOE) were built to assess the health risks of benzopyrene, aflatoxin B1, and heavy metals in edible vegetable oils. And then an early warning model of food oil safety risk was established by using analytic hierarchy process (AHP) and back propagation (BP) neural network. According to the national standards and sampling data, the risk early warning model determined eight evaluation indexes of vegetable oil quality and safety, and calculated the risk value of chemical pollution in vegetable oil by Entropy Weight-Analytic Hierarchy Process(EW-AHP). The data of eight evaluation indexes were taken as the input of the model, and the comprehensive risk value was taken as the output of the model. The model learning process was carried out, and four algorithms, namely Support Vector Machine (SVM), Random Forest (RF), Radial Basis Function Neural Network (RBF) and Neural Network (BP), were selected for construction and comparison of model. The daily sampling data of chemical hazard factors in edible vegetable oil were quantified into specific chemical hazard risk levels. Thus the goal of predicting chemical hazard levels was achieved in edible vegetable oil. This study provides targeted reference suggestions for the safety supervision of edible vegetable oil, so as to improve the efficiency of supervision and ensure the consumption safety of edible vegetable oil. • Quantitative risk assessment models for benzopyrene, aflatoxin-B1 and heavy metals in edible vegetable oils (EVO) in China. • Early warning model based on machine learning for EVO. • Analyses of exposure and health risk for EVO. [ABSTRACT FROM AUTHOR]