Abstract: In food safety risk management, accurately locating risk points is of critical significance to food safety risk assessment and prewarning, since it helps solve food safety problems from the source. Recently, with the development of informationization in the food industry, a large amount of food safety data accumulated during raw material production, food processing, storage and transportation and, sample inspection urgently need to be developed and utilized. However, there exist lots of deficiencies in the existing food safety risk prewarning system, such as rough risk measurement, low data utilization rate, and high labor cost. Therefore, in this paper, we sorted the food safety data and described the data features for subsequent processing. Meanwhile, in order to take full advantages of the data features of large quantity and high dimension, the combination of prior risk probability and fuzzy hierarchy partition was employed to calculate fuzzy comprehensive risk values based on various attributes for use as the expected output of a predictive model that can predict and validate risk values, generated using light gradient boosting machine (LightGBM) combined with experts’ modification operations. Finally, data on meat products and aquatic products were used to illustrate how to use this method, and its superiority and reasonability were validated. The risk analysis results in this paper, including the risk values and attribute importance distribution, can provide decision makers with valuable information.

Key words: food safety; risk prewarning; fuzzy hierarchy partition; gradient boosting decision tree; light gradient boosting machine