부산대학교 도서관

This paper presents a comparative analysis of five classification algorithms: k-nearest neighbor (k-NN), Support Vector Machine (SVM), Classification and Regression Trees (CART), Naive Bayes (NB) and Random Forest (RF) for recognition of three types of food ingested: solid, liquid and pasty, given that the use of these algorithms can optimize and assist in medical decision making. To achieve this goal, data on mandibular movements from the food intake process and the chewing time of 23 volunteers were captured. This data was captured through a noninvasive device. For training and validation of the algorithms, leave-one-out cross validation was used. As a result, the SVM algorithm performed better with a final accuracy of 89.8 %.