Research on Risk Prediction of Condiments Based on Gray Correlation Analysis – Deep Neural Networks

This is the authors' abstract. We don't add key points for this paper.

• Data mining and analysis based on big data of spice samples in China. • A risk assessment approach for spices was constructed through gray correlation analysis. • A risk prediction model for spices was constructed to enable risk assessment. • The methodology allows for a comprehensive risk assessment of spices. Food safety is directly related to the health of the public, and the safety of condiments is also of great significance. In this study, a risk assessment model for condiments based on gray correlation analysis was established by using publicly available sampling data of soy sauce and vinegar in China. Risk indicator screening and data preprocessing were performed first, and the weight of each indicator was calculated by gray correlation analysis to formulate a comprehensive risk value label. Then, three machine learning models, Deep Neural Network (DNN), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost), were used to predict the comprehensive risk values. Finally, the fuzzy synthesis analysis was utilized to classify the risk level of the composite risk value. In this study, based on the analysis of 282 sets of soy sauce and 704 sets of vinegar samples, the trained DNN model has the optimal prediction performance, which can basically predict the comprehensive risk value and risk level of a sample by inputting the detection indexes of that sample. This method can provide a more reasonable basis for relevant departments to formulate risk control strategies.