This study presents the implementation of an approach to classifying school dropout risk among secondary and high school students. By leveraging knowledge graph embeddings and machine learning algorithms, we aim to enhance classification accuracy and interpretability. We constructed a knowledge graph from a dataset of 1,830 students in a real scenario of a public educational institution; the records were composed of demographic and academic features. We trained the ComplEx model to generate embeddings. We subsequently used these embeddings in clustering and classification tasks via K-means, XGBoost, and Random Forest. The XGBoost classifier attained an F-score of 0.63 and an accuracy of 0.82, and the Random Forest classifier reached an F-score of 0.55 and an accuracy of 0.83. This integration of knowledge graph embeddings improved the model’s performance concerning baseline prediction, providing more precise insights into the factors influencing dropout risk. This approach contributes to a deeper understanding of the most influential factors in student dropout for the educational institution under study, supporting efforts to detect early dropout risk and success in educational settings.