Teachers who are aware of their students' strengths and weakness can tailor their teaching methodologies to meet the challenging students efficiently for better results. This helps them to identify any potential learning challenges at an early stage leading to improved academic performance and success ratio. This also fosters a learning environment where students feel motivated and valued to excel in their respective fields. This study offers a robust adaptive feedback system tailored for Learning Management System leveraging instance level explorations, helping teachers to find the specific instance affecting the learner's learning outcome. The proposed system can also be utilized by the institutions where the outcome-based education system has been adopted. The study includes Stacking, Capsule Network, SVM, Random Forest, Decision Tree, and KNN for experiments. Stacking achieved the highest accuracy of 76.70% while SVM demonstrated the highest precision of 0.78 showing the effectiveness of ensemble learning techniques. The primary objective of this endeavor is to elevate automated assessment to provide precise and meaningful feedback, enhancing the educational experience for tertiary students through the integration of technology and pedagogical concepts. The learning feedback has been made available via a user-friendly webserver at: https://khan-learning-feedback.streamlit.app/ .