Prediction of difficulty for video-based learning materials : an initial approach

This work is intended to have an initial approach towards the video-based learning materials difficulty prediction using machine learning techniques. First, we address the possible features that can be extracted from the video resources, propose a scheme of 7 domains of features that a video can have and extract the domains' features. After that, we introduce a data recollection platform that is used to obtain video class labels. We later treat this problem as a binary class problem using majority vote and use SMOTE to address class imbalance. We then use supervised learning algorithms to build difficulty prediction models. Later we perform experiments to determine the best feature subset and best algorithm to solve the problem. Finally, we obtain a prediction model using Random Forest which has an average accuracy of 90.3%.

• Autores: Venegas Bernal, Tomás Felipe; Lovera Lozano, Juan Manuel Alberto
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2018
• Institución: Universidad de los Andes
• Repositorio: Séneca: repositorio Uniandes
• Idioma: eng
• OAI Identifier: oai:repositorio.uniandes.edu.co:1992/45204
• Acceso en línea: http://hdl.handle.net/1992/45204
• Palabra clave: Aprendizaje automático (Inteligencia artificial); Inteligencia artificial; Algoritmos (Computadores); Ingeniería
• Rights: openAccess
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/