In the higher education (HE) sector, balancing the number of students enrolled and those who pass out is a big challenge. This issue leads to the loss of potential talent and negatively affects higher education institutes in terms of finance and academics. Student dropout is a multi-factor related problem that needs a contemporary approach to identify the main factors for predicting the student who has the chance to drop out. Nowadays, applying various Machine Learning (ML) techniques in the Education sector has gained much attention from educators and education administrators. The principal research objective of the paper is to develop a machine learning approach to predict the possibility of academic failure of a student in the higher education path. The authors define academic failure as a dropout in the middle of a course and academic success as completing the course within a particular duration. So, from the ML‘s point of view, the study deals with classification problems, specifically binary classification. Through this study, the authors try to find the ML models for the issue by comparing the performance of different state-of-the-art algorithms based on the Enrolled Students (ES) dataset. The authors use a stacking model that uses a multilayer perceptron as a meta-classifier, Random forest and Gradient Boosting as Base Classifier, which gave better results than classical algorithms like Logistic Regression (LR), Naive Bayes (NB), Decision Tree (DT), Random Forest (RF), Multilayer Perceptron (MLP). The developed stacking model gave us the best accuracy 87%. This model also obtains good scores for other performance metrics like sensitivity, specificity, ROC-AUC and Kappa Statistics.