AI Based Student Absenteeism Prediction with Personalized Recommendations: A Study in Batticaloa Manmunai West Zone

Abstract

Student absenteeism is a growing issue, especially in rural and under-resourced areas like the Manmunai West Educational Zone in Bat ticaloa, Sri Lanka. Poor attendance has a strong impact on students’ academic performance and long term success. Factors such as family financial issues, school conditions, and personal health or motivation problems contribute to this challenge. The research identifies the most important factors influencing ab senteeism. As well as this study aims to develop an AI-based system to predict absenteeism risk among secondary school students and provide personalized recommendations to help reduce that risk. A struc tured 29-question Likert-scale questionnaire was ad ministered to 116 students, covering family, school, and student-related issues. Data were analyzed using SPSS and Python. Students were labeled into High, Medium, or Low absenteeism risk levels based on their average scores. A Random Forest classifier was trained to predict the risk level. LIME (Local Inter pretable Model-agnostic Explanations) was used to explain the model’s decisions. The model’s accuracy was validated using Support Vector Machine (SVM) and 5-fold cross-validation. A Streamlit-based web application was also developed for practical use. The Random Forest model achieved 96% accuracy in classifying absenteeism risk. Important influencing factors included low family income, parental support, school safety, and student health issues. The find ings from SPSS mean analysis were consistent with Python model outputs. LIME effectively explained the predictions, helping educators understand the root causes of absenteeism. The proposed AI-based model is accurate, interpretable, and practical for identifying students at risk of absenteeism. The web based tool allows schools and educational offices to take early action with personalized interventions. This system has the potential to improve attendance and learning outcomes in underprivileged regions like Batticaloa’s Manmunai West Zone. Future re search can improve accuracy further by expanding the dataset and including real-time attendance data.