Evaluating SVM and Soft Voting Ensemble Methods for Classifying Stress Levels Among University Students

This study investigates the classification of university students' stress levels using anonymous text posts collected from online platforms where students commonly express their emotions. The dataset was acquired through a web scraping process and manually labelled into three categories: no stress, mild stress, and high stress. Two machine learning approaches were implemented and compared: a standalone Support Vector Machine (SVM) model and a Soft Voting Ensemble that integrates SVC, Logistic Regression, and Random Forest classifiers. Model performance was evaluated based on accuracy metrics on both training and validation datasets. The results indicate that the Soft Voting Ensemble outperforms the SVM model, demonstrating greater effectiveness in multi-class text classification tasks involving unstructured and emotionally nuanced student-generated content.