Evaluation of Facial Attractiveness after Undergoing Rhinoplasty Using Tree-based and Regression Methods

Associations between geometric features of a human face and facial attractiveness level is important for planning of facial aesthetic surgeries including rhinoplasty, but is complex enough and remains still unclear. Going further, facial attractiveness is also dependent on currently expressed facial emotions, therefore an area of facial plastic surgery needs a reliable way how to classify each patient’s facial image into one of the facial emotions.To address both of the challenges, we performed regression trees- based analysis in order to realize which changes of geometric features of a human face increase its attractiveness level after undergoing rhinoplasty. Multivariate linear regression was applied to quantify effect sizes of the features’ changes.Naïve Bayes classifies, classification trees, random forests, support vector machines and neural networks, respectively, were learned to classify facial images into facial emotions.Enlargement of both nasofrontal and nasolabial angles increase statistically facial attractiveness after undergoing the rhinoplasty, as both tree-based and regression methods showed. Classification accuracy of the neural networks exceeded accuracies of other machine-learning methods.The applied machine-learning methods uncovered some significant facial geometric features increasing facial attractiveness after the rhinoplasty undergoing as well as possibility to classify facial images into facial emotions.