Identifiering av elevers stödbehov med hjälp av maskininlärning

This study investigates how machine-learning techniques can be employed to identify students at risk of academic underperformance. The motivation for this study is to help teachers distribute their resources more equitably and to give students from different contexts a similar chance at success. Two datasets are used: a sample of Portuguese students from the UCI ML Repository and a subset of Swedish students from the PISA 2022 survey. These datasets are preprocessed by employing strategies such as data balancing, feature transformation, and regularization. Machine learning is performed using linear and logistic regression models, due to their interpretability Results show that logistic regression with L2 regularization achieves the highest predictive performance on both datasets, with F-scores of 0.633 and 0.772, respectively. Key predictors include prior academic failures, digital resource usage, and aspirations for further education. The findings indicate that interpretable models can provide actionable insights into student risk classification, supporting more targeted educational interventions.