PREMUCC: Design proposal for a microservices architecture for the development of a context-aware dropout prediction system

Laury Arenales, Joseph E. Gonzalez, Juan Jose Saldana-BarriosInge 2024 · DOI

Abstract

Introduction: This article presents the PREMUCC system, a proposal based on microservices to predict context-aware college dropout in higher education. Student attrition is a concerning issue addressed by this innovative solution, enabling personalized support and informed decision-making. Objective: The objective of the PREMUCC system is to anticipate university dropout through the use of machine learning techniques and a microservices architecture, providing personalized support to students and facilitating educational management. Method: The proposal for PREMUCC was based on analyzing dropout prediction systems and microservices architectures. Technologies and data analysis processes were identified to ensure system efficiency and accuracy. Results: The PREMUCC system integrates six microservices encompassing user authentication, previous studies analysis, psychological testing, grade tracking, financial evaluation, and the prediction system. Students access personalized information to improve their academic performance. Conclusions: The PREMUCC proposal represents a significant advancement in mitigating college dropout. Its focus on microservices and modern technologies enables closer tracking of each student's context, supporting informed decision-making. Its implementation could enhance the educational system, saving resources, time, and effort.

Introduction

Student attrition is a global crisis in higher education, particularly acute in technical and engineering fields. In the Republic of Panama, while the national dropout rate is approximately 22.7%, the figures at the Technological University of Panama (UTP) are significantly more alarming. Between 2016 and 2020, engineering programs experienced a staggering 58.29% dropout rate. In specific disciplines like Computer Systems Engineering, this figure reached as high as 82.42%.

Dropout rates by faculty at UTP (2016-2020)

Existing prediction systems often fail because they are too narrow, typically focusing solely on academic grades while ignoring the broader "context"—socio-economic, psychological, and personal factors—that drive a student to leave. Furthermore, many current tools are built as monolithic applications, making them difficult to scale or integrate with modern mobile and cloud environments.

Proposed Architecture

To address these challenges, the PREMUCC system (a Spanish acronym for a context-aware dropout prediction system) utilizes a four-layer microservices architecture designed for scalability and fault tolerance.

The four-layer cloud architecture of PREMUCC

The architecture is organized into the following layers:

  1. User Layer: A multi-terminal front-end (web and mobile) for students, faculty, and administrators.
  2. Processing Layer: An API Gateway using REST protocols to manage communication between the UI and the backend services.
  3. Microservices Layer: The core logic, split into six independent, containerized services (Authentication, Previous Studies, Psychological Test, Score Tracking, Economic Support, and the Prediction System).
  4. Data Layer: A decentralized storage approach where each microservice maintains its own database to ensure independent operation.

Key Insights and System Components

The innovation of PREMUCC lies in its "context-aware" approach. By treating psychological profiles and financial status as first-class citizens within the microservices layer, the system can synthesize a more holistic risk profile. For example, Service E (Economic Support) evaluates scholarship status, while Service C (Psychological Test) integrates entrance exam results to identify potential mental health or motivation hurdles.

This decoupling ensures that the failure of one service does not crash the entire system. Moreover, the mobile-first design supports "ubiquitous learning" (u-learning), allowing the university to send real-time "nudges" or interventions directly to a student's device based on the machine learning engine's risk assessments.

Conclusion

The PREMUCC framework represents a validated design that bridges the gap between high-level architectural needs and the specific socio-academic requirements of higher education. While currently a design proposal, its comparative analysis against existing systems shows it is uniquely positioned to offer cloud availability, context awareness, and predictive capabilities simultaneously. Future work will focus on full implementation and the selection of optimal machine learning algorithms to maximize the accuracy of dropout interventions.

Summary updated

Abstract

Introduction: This article presents the PREMUCC system, a proposal based on microservices to predict context-aware college dropout in higher education. Student attrition is a concerning issue addressed by this innovative solution, enabling personalized support and informed decision-making. Objective: The objective of the PREMUCC system is to anticipate university dropout through the use of machine learning techniques and a microservices architecture, providing personalized support to students and facilitating educational management. Method: The proposal for PREMUCC was based on analyzing dropout prediction systems and microservices architectures. Technologies and data analysis processes were identified to ensure system efficiency and accuracy. Results: The PREMUCC system integrates six microservices encompassing user authentication, previous studies analysis, psychological testing, grade tracking, financial evaluation, and the prediction system. Students access personalized information to improve their academic performance. Conclusions: The PREMUCC proposal represents a significant advancement in mitigating college dropout. Its focus on microservices and modern technologies enables closer tracking of each student's context, supporting informed decision-making. Its implementation could enhance the educational system, saving resources, time, and effort. It is proposed to strengthen the architecture and evaluate its effectiveness in real higher education settings.
BibTeX
@inproceedings{arenales2024premucc,
  author = {Laury Arenales and Joseph E. Gonzalez and Juan Jose Saldana-Barrios},
  title = {{PREMUCC: Design proposal for a microservices architecture for the development of a context-aware dropout prediction system}},
  booktitle = {Inge},
  year = {2024},
  doi = {10.17981/ingecuc.20.2.2024.09},
}
Figure