Data science for a new generation of tutors: building an academic-guidance system based on dropout and grades prediction

Abstract

This work is part of an innovative educational project which aim is to create a tool to help tutors offer more personalised and proactive guidance to the students. An analysis of the performance of different Machine Learning techniques for dropout intention prediction is presented. The approach of using Recommender Systems for final grade prediction and course ranking creation has been also assessed. Visualizations which help in the interpretation of the obtained results have been developed and a design for the tutoring tool has been outlined. The research has been performed using data from the degree studies in Law, Computer Science and Mathematics of Universitat de Barcelona.