This study conducts a comparative analysis of machine learning and deep learning algorithms to predict university dropout rates. Using a dataset from a university, specifically Tecnológico de Monterrey in Mexico, we establish a interactive pipeline base on CRISP-DM and several platforms and techniques to evaluate algorithms including TabNet, LSTM, Logistic Regression, and SVM. The goal is to identify the most effective model in accurately predicting potential dropout cases. The results show that the proposal pipeline contributes to algorithms achieving the highest accuracy, for instance TabNet reaching 97% of accuracy. This indicates its effectiveness in recognizing patterns associated with dropout risks linked to academic performance and economic resources. The comparative analysis highlights the strengths and weaknesses of each algorithm in handling the complexities of university dropout prediction, providing insights for future research and application in educational settings to student dropout prediction.