Cross-Correlation and Causality between Physicochemical Properties and Wine Quality

The physicochemical properties and wine quality is a high dimensional complex network, For evaluating wine quality, an accurate examination for physicochemical properties of cross-correlation and causality is critical. In this study, Multi-Fractal Detrended Fluctuation Cross-Correlation Analysis (MF-DCCA) and Transfer Entropy (TE) coupled with the complex network analysis for wine quality investigating by determining the nonlinear directional relationship between the physicochemical properties. By MF-DCCA complexity parameters regard to cross-correlations are retrieved. All the physicochemical properties sequence indicates significant and high multifractality and persistent behavior for red wine and white wine. An obvious causal relationship and bidirectional spillover effect is observed by TE between physicochemical properties and wine quality. Dynamic information flows show the obvious erratic dynamic patterns respectively for red wine and white wine. And the predictability of the wine quality is verified by physicochemical properties. Our work broadens the cognition of wine taste for the wine industry and researchers, as well as providing a deep cognition relationship between physicochemical properties and wine quality. Also deepens understanding of interpretability and uncertainty for machine learning model