Risk early warning of food safety using novel long short-term memory neural network integrating sum product based analytic hierarchy process

Purpose Food safety risk brought by environmental pollution seriously threatens human health and affects national economic and social development. In particular, heavy metal pollution and nutrient deficiency have caused regional diseases. Thus, the purpose of this paper is to present a risk early warning method of food safety considering environmental and nutritional factors. Design/methodology/approach A novel risk early warning modelling method based on the long short-term memory (LSTM) neural network integrating sum product based analytic hierarchy process (AHP-SP) is proposed. The data fuzzification method is adopted to overcome the uncertainty of food safety detection data and the processed data are viewed as the input of the LSTM. The AHP-SP method is used to fuse the risk of detection data and the obtained risk values are viewed as the expected output of the LSTM. Finally, the proposed method is applied on one group of sterilized milk data from a food detection agency in China. Findings The experimental results show that compared with the back propagation and the radial basis function neural networks, the proposed method has higher accuracy in predicting the development trend of food safety risk. Moreover, the causal factors of the risk can be figured out through the predicted results. Originality/value The proposed modelling method can achieve accurate prediction and early warning of food safety risk, and provide decision-making basis for the relevant departments to formulate targeted risk prevention and control measures, thereby avoiding food safety incidents caused by environmental pollution or nutritional deficiency.