Risk prediction model for food safety based on improved random forest integrating virtual sample

Food safety has a severe impact on the world economy and global health, and improving the prediction accuracy and prevention ability of food safety risks and hazards protection is of great significance to the society sustainable development. However, the sample size of most food data is very small, resulting in insufficient training sample data for data-driven risk prediction models in low prediction accuracy and inability to provide effective prevention and control measures. Therefore, this paper proposes a food safety risk prediction model based on an improved random forest prediction method integrating Monte Carlo algorithm to protect and reduce the food safety risk. The Monte Carlo algorithm can expand the small sample data for obtaining the generated virtual sample. Then the extended input of a random forest method to construct the novel risk prediction model of food safety for predicting food risk levels and ensuring personnel safety. Finally, the accuracy and effectiveness of the proposed method is verified by the sterilized milk data. The experimental results show that the novel risk prediction model out-performs state-of-the art, which has stronger generalization ability and higher prediction accuracy for realizing effective risk early warning. Moreover, the proposed model can provide decision assistance and technical support for relevant departments to prevent and control in advance, and reduce the occurrence of food risk events.