Learning analytics always has trouble with student engagement and academic performance due to students dropping out. Predictive algorithms that have been around for a long time can find students who are likely to drop out, but they can't say the reason why the students dropout. Because there isn't enough transparency, teachers and academic decision-makers have a hard time understanding the main risk factors that lead to students dropping out. This study presents an Explainable Artificial Intelligence (XAI)-based early warning and intervention framework to help with the problems that come with predicting risk-stratified student attrition. Ensemble machine learning models such as Random Forest, Gradient Boosting, and Stacking classifiers can be employed to predict the likelihood of an individual’s dropout based on academic, behavioural, and socioeconomic variables. SHapley Additive exPlanations (SHAP) are used to explain each prediction in a clear way at the feature level. A systematic intervention framework connects each level of risk to certain academic and mentoring programs. Students are put into groups based on how likely they are to take risks: low, medium, or high. frame work can support to link risk level to academic and monitoring approaches. The interventions are planned to allow for changes and risk assessments that can adapt through a system of constant monitoring. The model does well on Accuracy, F1-score, Recall, and ROC-AUC tests, and it's still easy to understand. This study presented a framework for learning analytics that converts predictive analytics into a proactive and comprehensible decision-support system.