Wine And Liquor Research Based On Correlation And Machine Learning Algorithms

The physicochemical properties of wine reflect the composition of the wine,the composition of the wine determines the quality of the wine.The quality of the wine is evaluated by experts and lacks objective measurement standards.How to accurately analyze the correlation between the physicochemical properties and quality of wine and predict the quality objectively and effectively is of great significance to the improvement of wine quality.However,the relationship between the physicochemical properties and quality of wine is very complicated and not fully understood.Therefore,this paper introduces correlation theory and machine learning methods to study wine and liquor based on correlation theory and machine learning algorithms.When studying the correlation of wine,this paper introduces Pearson correlation coefficient,gray correlation analysis,and multifractal de-trended cross-correlation analysis methods to systematically analyze the correlation between the physicochemical properties and quality of the public wine dataset.The Pearson correlation results show that there is a linear correlation between the physicochemical properties and the quality;the gray correlation analysis shows that the physicochemical properties and the quality have a medium or higher strength coupling effect;the multifractal de-trended crosscorrelation analysis shows that there is a weak long-range cross-correlation between the physicochemical properties and the quality of wine.On the basis of correlation analysis,this article further models wine quality,and introduces 14 machine learning classification algorithms to predict wine quality,that is,K nearest neighbors,support vector machines,naive Bayes,decision trees,random forests,Adaptive Boosting(Ada Boost),Gradient Boosting Decision Tree(GBDT),e Xtreme Gradient Boosting machine(XGBoost),Light Gradient Boosting Machine(Light GBM),Category boosting(Catboost),shallow neural network(SNN),deep neural network(DNN),long short-term memory network(LSTM),Convolutional Neural Network-Light Gradient Boosting Machine(CNN-Light GBM),and compares the evaluation indicators of the classification algorithms.Machine learning algorithms can predict wine classification very well,but the classification indicator performance of different algorithms is quite different,and the evaluation indicator of the same classification algorithm has different sensitivities to wine types.The best predictive performance in this article is the decision tree and the ensemble algorithm,and the area under the curve(AUC)values of them are all above 0.95.In the study of liquor,due to the lack of physical and chemical datasets,it is impossible to carry out the correlation analysis and quality prediction modeling of liquor.This article introduces two spatial autocorrelation indicators,the global Moran’s I index and the local Moran’s I index,which are cross-combined with the Baidu index,innovatively analyzed the spatial autocorrelation between the sales volume of liquor on Tmall and the place of production,and the correlation between the search volume of liquor brand keywords included in the Baidu Index and the place of production.Liquor sales data does not have global spatial autocorrelation,and there is spatial autocorrelation locally between them,and sales are shown as local aggregation;Baidu Index liquor brand search volume and liquor production area have a strong spatial correlation,which provides evidence for the local spatial aggregation of sales.