Academic Advisors Optimization

Overview

Managing the assignment of academic advisors (garanti) in a university setting is a complex task involving numerous organizational factors and large datasets. Manual processes are often time-consuming and rely on informal criteria, which can limit flexibility and efficiency.

This project presents an automated system based on Answer Set Programming (ASP), a declarative logic programming paradigm particularly suited for combinatorial optimization problems. By integrating Python with the Clingo solver, the system can process heterogeneous input data to automatically identify the appropriate academic advisors according to specific institutional criteria.

Motivation

The University of Parma currently handles advisor assignments manually, resulting in lengthy and inflexible procedures. Automating this process can improve administrative efficiency, ensure consistency, and enable real-time adaptability to changing requirements.

Methodology

• Answer Set Programming (ASP): Used to declaratively model the problem, expressing constraints and optimization criteria naturally.
• Clingo Solver: An open-source ASP solver integrated with Python to manage data preprocessing, ground fact generation, and optimization solving.
• Input Data: Collected from multiple sources within the university, including course coverage, teaching staff lists, and enrollment data.

Implementation

The solution follows a pipeline approach:

1. Data normalization and preprocessing using Python.
2. Generation of ground facts representing the problem state.
3. Execution of ASP solving via Clingo APIs.
4. Post-processing results into a user-friendly tabular format.

Specific constraints modeled include:

• Assigning the course president automatically as advisor.
• Maximizing the number of tenured professors serving as advisors.
• Minimizing the involvement of temporary contract professors and researchers.
• Enforcing limits on the number of advisors per course and their qualifications.

Weak constraints with prioritized optimization levels allow nuanced preference management.

Results

• The system efficiently generates valid advisor assignments respecting all constraints.
• Optimal solutions are found quickly for single or paired courses.
• For larger datasets, solution times vary, but the solver produces correct and near-optimal results within practical time limits.
• Introducing a timeout ensures responsiveness even in complex scenarios.

Conclusion

The project demonstrates the effectiveness of ASP and Clingo in automating a complex academic administrative process. The approach successfully handles real-world data and constraints, paving the way for smarter university management tools.

Future enhancements include supporting weighted advisor assignments across multiple courses, developing a centralized, real-time accessible database, and creating a user-friendly interface for academic staff.