Support Vector Machine Method for Predicting Children's Emotions

Abstract

Emotion is an intense feeling that is directed at someone or something that can trigger to take an action or expression from inside or outside themself. In everyday life, it is important to be able to understand and predict the emotional state of a person. One of the emotions of a person can be known through facial expressions, but sometimes a person can manipulate it by controlling his facial expressions. Therefore, we need a system that can predict children's emotions based on changes in their body condition. In this study, two sensors were used, namely the GSR sensor and the heart rate sensor. The data obtained from each sensor is processed using a microcomputer or raspberry pi. Furthermore, the sensor output will be processed first using the Support Vector Machine (SVM) method by comparing 2 pieces of data, namely normalized data and unnormalized data based on linear, polynomial, RBF, and sigmoid kernels, so that a model will be obtained and will be used as a reference to predict 3 types of emotions namely happy, sad, and angry. From this research, the best model is obtained when using linear and sigmoid kernels with normalized data types. This is proven when using normalized data with linear and sigmoid kernels, there are only 18 incorrect data with an accuracy rate of 86.67%. This research is expected to help parents and teachers to predict emotions, especially in children.