Nowadays, the role of a tutor is more important than ever to prevent students dropout and improve their academic performance. This work proposes a data-driven system to extract relevant information hidden in the student academic data and, thus, help tutors to offer their pupils a more proactive personal guidance. In particular, our system, based on machine learning techniques, makes predictions of dropout intention and courses grades of students, as well as personalized course recommendations. Moreover, we present different visualizations which help in the interpretation of the results. In the experimental validation, we show that the system obtains promising results with data from the degree studies in Law, Computer Science and Mathematics of the Universitat de Barcelona.

No abstract

This paper aims, first, to describe the fundamental characteristics and workings of the AgentGeom artificial tutorial system, which is designed to help students develop knowledge and skills related to problem solving, mathematical proof in geometry, and the use of mathematical language. Following this, we indicate the manner in which a secondary school student can appropriate these abilities through interactions with the system. Our system uses strategic messages of the agent tutor in an argumentative process that collaborates with a student in the construction of a proof.

No abstract

In many classification problems, neighbor data labels have inherent sequential relationships in spite of their values. Sequential learning algorithms take benefit of these relationships in order to improve generalization.In this paper, we revise the Multi-Scale Sequential Learning approach (MSSL) for applying it in the multi-class case (MMSSL). We introduce the Error-Correcting Output Codes (ECOC) framework in the MSSL classifiers and propose a formulation for calculating confidence maps from the margins of the base classifiers. In addition, we propose a MMSSL compression approach which reduces the number of features in the extended data set without a loss in performance. The proposed methods are tested on several databases, showing significant performance improvement compared to classical approaches.