Research On Classification Model Of Fermented Milk Quality Control Based On Data Mining

Fermented milk refers to a curd-like product made by lactic acid fermentation by adding lactic acid bacteria to milk.Because of its supple taste and good health care value,it is highly praised by Chinese consumers.Currently,artificial sensory evaluation is generally used to judge whether the quality of fermented milk meets the standard.However,this analysis method has obvious shortcomings and shortcomings.For example,the selection and training of evaluators is time-consuming and laborious,qualified evaluators are difficult to convene,reproducibility is poor,and quantification is difficult.In addition,the artificial sensory evaluation is highly subjective,and is easily affected by the personal,physical and mental condition of the appraiser.In order to solve the above problems,the experimental team collected the sample data of fermented milk provided by the electronic sensory analysis system,combined with artificial sensory,and marked the quality of fermented milk in each process.The key control points for fermented milk production include raw milk,semi-finished liquid milk(after sterilization),semi-finished fermented milk(after fermentation),and finished fermented milk in storage.According to the analysis of quality control elements,only the sense of smell is involved in the four-point quality assessment.Therefore,the experiment starts with the sense of smell,and based on the electronic nose data of different control points,the integrated classification model Ada Boost?NB and Bagging?NB are constructed with the help of Naive Bayes(NB).After comparison,the average performance index of the model based on Ada Boost?NB at the four control points is better than that of a single model such as SVM,NB,and KNN.It is also concluded that linear discriminant analysis(LDA)has a better dimensionality reduction effect than principal component analysis(PCA).In order to further enhance the integrity and rigor of the quality control ideas,sensory data such as color,acidity,texture and so on were later integrated.Without changing the data distribution,standardization is used to complete preprocessing,and a heterogeneous ensemble classification model based on Stacking is proposed,which verifies that multi-sensory data fusion can also improve classification performance to a certain extent.Therefore,the integrated model can be used for the quality classification of these control points.Experimental results confirm that model fusion can make up for the shortcomings of a single model and has a better classification effect.In this way,the purpose of providing a better theoretical basis for the quality control of fermented milk is better achieved.In order for the research results to enter the production environment in the future,this paper also completed the database design and page implementation of the quality control classification APP.