Accurately predicting how well students will perform is critical, as it hinges on numerous factors. Machine learning and data mining tools offer powerful solutions for this task. This paper delves into the use of XGBoost, Random Forest, Gradient Boosting Machines, Support Vector Machines, and Convolutional Neural Networks for predicting student success. Three datasets from various levels (school, college, and online) with diverse input parameters are used to compare the effectiveness of these five techniques. The paper not only presents comparative results but also explores how their performance changes with different tuning parameters. By evaluating their capabilities across three distinct datasets, the study reveals that Decision tree and CNN with Bidirectional Long Short-Term Memory outperform the other three methods.