It is essential to be able to predict the academic success of students in order to improve educational outcomes for students. Noting the student's academic performance is a multi-faceted problem and one that has great economic value for educational institutions and students. This study seeks to analyze whether cloud machine learning tools could assist in the measurement of student success in terms of prediction models. We apply several supervised learning algorithms such as regression, decision tree and neural network models in an attempt to build different student outcome prediction models like the students’ final grade score, GPAs, and the percentage of the students who complete the program. This research focuses on addressing the problem of determining the potential students’ academic success through cloud-based machine learning predictive models that would enable the institutions to implement timely measures for students who are about to fail and at-risk students.

The research methodology comprises the activity concerning, the collection and preprocessing of old data ranging from “10,000 students during 5 years period” as well as demographic, academic, and behavioral data. The relevant characteristics that affected student outcomes were determined by means of feature selection techniques which improved the performance of the model by placing emphasis on relevant predictors. For the building of predictive models, a variety of machine learning algorithms including linear regression, decision trees and neural networks were utilized, where models were trained and tested on cloud platforms for their scalability and computational efficiency. In terms of the methodology we gathered and studied a set of comprehensive datasets pertaining to students’ academic records, demographics and other related variables. The models were assessed against performance indicators including accuracy, precision, recall and F1-score, and the results confirmed that “neural networks were found to outperform the other algorithms with respect to the accuracy of predicting the final grades”. These models were then transferred on cloud based solutions avoiding dependency on local IT infrastructure or resources.

The principal outcome of this investigation is that the developed models proved “a good accuracy in predicting success of students” and especially pinpointed for students' success such factors as ”previous schooling achievements, participation, and activity in out-of-school hobbies”.

These last findings are highly relevant to educational organizations. Established models may assist in addressing students’ concerns before they become persistent issues through means such as strategies, academic counseling, tutor sessions, and individualized methods. This research supports the broader literature in educational data science by showing a practical model for increasing student performance and retention in the context of a cloud-based business model, showing a use case. These implications have broader relevance, such as for decision makers and managers who want to enhance specific educational goals and the use of resources in an institution.

Keywords: Education, machine learning, cloud computing, student success, academic performance, predictive analytics.