Red wine has become an integral part of people’s lives and culture. Modeling the red wine quality is crucial. We propose a new framework to predict the red wine quality ratings. MF-DCCA was utilized to quantitatively investigate the cross-correlation between quality and physicochemical data. The long-range correlations importance was ranked. We compared two machine learning algorithms with other common algorithms implemented on the red wine data set, which was taken from UC Irvine Machine Learning Repository to ensure the reliability and performance. These data sets contain 1599 instances for red wine with 11 features of physicochemical data. Our model has better performance than other results.