Application of SVM in the Classification of Data Obtained from Facial Expressions

Objective: To classify a set of data captured from facial expressions when exposed to marketing stimuli in the laboratory, in order to predict consumer preferences and purchase decisions. Methodology: Data was recorded from ten subjects while they randomly watched four videos obtained from the web about advertising of antiperspirant sticks for men. FaceReader software was used for the acquisition of the facial images and for the creation of the dataset in CSV format. Results: Data classification is performed using the support vector machine and decision tree algorithms, with both balanced and unbalanced data, to observe improvements in prediction results. Based on the achieved results, it is observed that the SVM model with balanced data provides better predictions.