Application Research On The LightGBM Model Integrated With INMCSA For Parameter Adjustment In Food Safety Risk Analysis And Early Warning

Food safety is a key issue related to people's livelihood and national stability in the world.In recent years,the food safety problems are of common occurrences,which cause lots of damages to the world economy.Food safety risk analysis and early warning,with a great theoretical significance and practical value,play key roles to accurately locate the risk points,in-time make response strategies,and further,resolve the problems from the source.However,a wide variety of food,complex schedules of process,storage,transport,and various ingredient components,make it more difficult to promote this work.Moreover,with the intellectualization and informationization development of food safety supervision and administration,a large amount of valuable data have been accumulated.This paper,based on the food safety data and focusing on the food safety risk analysis and early warning methods,conducts a preliminary research and makes some progresses.In detail,after the food safety data is classified in terms of feature forms,the comprehensive risk value is calculated as training data risk labels for the output of prediction model(Light GBM)by the combination method of prior risk probability and fuzzy hierarchy.As the data scale increases,the Light GBM's adjustable parameters take a far greater impact on prediction accuracy and efficiency.Therefore,the Improved Novel Modified Cuckoo Search Algorithm(INMCSA)is used to assist the parameter adjustment in training process.The experiments demonstrated that,for the specific data scale,Light GBM outperforms compared methods on risk value regression prediction.Additionally,the INMCSA performs better than other parameter adjustment methods.In order to decrease the errors of comprehensive risk computing,the expert intervention strategy is used to correct the prediction model.On basis of the above,the food safety risk analysis and early warning system is introduced,implementing some data information display and risk analysis and prediction.It promotes the process of food safety risk analysis and early warning problem,and is of a great significance for food safety intellectualization supervision.