Leveraging Machine Learning, Explainable AI, and Open-Source LLMs for Students’ Placement Prediction

Students’ placement prediction is a significant challenge in academia. It requires robust machine learning techniques to evaluate academic performance and related parameters influencing employability. This research explores modelling students’ three semester examination performance data by integrating machine learning, explainable artificial intelligence (AI) and open source large language models (LLMs). The supervised machine learning techniques applied to classify students’ placement outcomes. The performance evaluation of classifiers used to predict placement. The techniques include Logistic Regression, Decision Tree, Random Forest and Support Vector Machine. The dataset for the present study designed by collecting MCA students’ performances in ‘In semester examination’ (ISE) and ‘End semester examination’ (ESE) of three semesters. The results present insights into various classifiers' performance and effectiveness in placement prediction. The model interpretability using explainable AI techniques added to research by leveraging SHAP (SHapley Additive Explanations) to determine which course scores contribute most to placement outcomes. Furthermore, natural language explanations of model predictions are generated by using open-source LLMs from Hugging Face’s Transformers library, specifically Meta’s LLaMA2 7B. The results show that explainability techniques combined with LLMs improve transparency in AI-driven student placement prediction solutions.