Comparison of machine learning algorithms for automatic assessment of performance in a virtual reality dental simulator

• Lucas H. Sallaberry Universidade de São Paulo - USP
• Romero Tori Universidade de São Paulo - USP
• Fatima L. S. Nunes Universidade de São Paulo - USP

Resumo

Virtual reality simulators for training of apprentices have been successfully developed for many different medical fields. However, establishing an objective and automatic method to assess the student’s performance is still difficult to achieve. In this study, we extracted several features from a dataset collected in a virtual reality simulator and compared the performance of different machine learning classification algorithms (Naive Bayes, Random Forests, Multi Layer Perceptrons and Support Vector Machine) and feature select/fusion algorithms (ReliefF and PCA). We found that the best results for each algorithm was usually achieved when combined with the ReliefF feature selection algorithm.

Palavras-chave: virtual reality, force feedback, Performance assessment, dental simulator