The development of a country is highly dependent on education and its technology. Sustainable development of once country influenced by educated person. In Ethiopia there are educational movement stages such as entrance exam which is the movement to join higher education. but to pass and achieve good result as well as joining to higher education influenced with factors such as social, environmental, economic, demographic, and schoolrelated, etc. influenced the success of individual‟s journey. Determine these causes for disruption of success is so important for improving individual achievement. To address these factors the study aims is predict students‟ national exam performance using machine learning techniques. The algorithms used for prediction models are Decision Tree (DT), Random Forest (RF), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and XGBoost (XGBoost). A total of 1149 data was collected from 1st and 2nd year Bahir dar, Debre tabor and Debre Markos University and Blue Nile College students through questionnaires in Google sheet form and direct hard copy form. The collected data are prepared using preprocessing techniques and appropriate features are selected for analysis. The models are trained with prepared data to predict whether the students are failing or pass by examining the contributing factors. The performance of the model is evaluated using performance metrics such as precision, recall, accuracy, and F-score to assess how effectively the model predicts the student‟s status. The results of the experiments show that XGBoost, which produced a value of 0.95.3%accurcy with grid search and Synthetic Minority Over-sampling Technique (SMOTE) balancing techniques, outperforms the other machine learning techniques. Finally, we used SHAP and LIME explanations on XGBoost models to understand how it works and gain insights into its predictions. This study provides valuable support to educators by helping them identify students who are facing academic challenges and gain a deeper understanding of their students' attitudes through behavioral observations. Adding more data collected from additional higher education and includes disable students data to improve the model performance mentioned for future investigator. Keyword: Balancing Techniques, Feature, Higher Education, Machine Learning, SMOTE.