Abstract:

The STEM degrees are fundamental for the advance of society, for this reason it is important that people that count with the profile for developing in them will be correctly oriented, in order to avoid early college dropout while they are studying a STEM career. The main reason for vocational orientation is guiding students to a successful career, but the classic vocational tests have the disadvantage of being excessively large and time consuming, with a duration of more than 45 minutes. In consequence, a shorter and reliable alternative is proposed. This paper presents an alternative method using a XGBoostClassifier algorithm trained with information extracted from biometrical records (electroencephalogram EEG and eye-tracking ET), belonging to 44 people with their answers from short traditional vocational test questions. We tested the XGBoostClassifier model with different hyperparameters and variations of our dataset, which achieved very good scores in the metric selected F1 score. We obtained results that showed that the combination of Machine Learning models with biometrical data could improve the vocational orientation.

Keywords:

XGBoostClassifier, Vocational Orientation, STEM, EEG, Eye-Tracking.