Early Detection of Emotional Issues in High School Students: A Machine Learning Approach

The primary objective of this project is to develop a predictive model for early emotional issue detection in high school students. By harnessing academic performance, attendance and discipline records, this research work seeks to detect early indicators of emotional issues, thereby empowering school administrators and educators to promptly identify and support students. At the heart of this novel approach lies the development of a machine learning model consisting of two essential components: a deep encoder and a classifier. The deep encoder performs feature extraction from time series data reflecting academic performance, by identifying and highlighting inherent patterns within these sequences. Such patterns hold critical information necessary for the early detection of emotional issues. The deep encoder’s output, known as embeddings, is then fed into the classifier for the classification process. This approach has proven remarkably successful, with the classifier achieving a remarkable 93% weighted F1 score on the test set. To achieve this result, we carefully developed a methodology to systematically explore the data, generate meaningful time series representations, and classify students’ academic performance in three classes, namely stable, monitor, or critical. This approach simplifies intervention efforts, leading to enhanced detection times.