ID: F202512091115 Tags: Verantwoording, Verslag, avans 2-1 LU2 File Link: Avans 2-2 LU2 Verantwoordingsverslag.pdf Status: File-Descriptor

File - VerantwoordingsVerslag Avans 2-1 LU2

Project: AI-Powered Elective Module Recommendation System for Avans University
Authors: Persoon 1 & Persoon 2
Date: 29/11/2025
Context: LU2, Year 2, Semester 1

Overview

This accountability report documents the development of a proof-of-concept AI recommendation system designed to help students select appropriate elective modules from Avans’ portfolio of 211 modules based on their values, interests, and goals.

Key Phases

Business Understanding

• Problem: Students struggle to find suitable modules among 211 options; often choose based on chance rather than personal fit
• Solution: Create a content-based recommendation system to streamline module discovery
• Stakeholders: Students, study advisors, module coordinators, curriculum management, ICT, and quality assurance

Data Preparation

• Dataset: 211 modules with descriptions, learning outcomes, and metadata
• Quality issues: Missing data, duplicate content fields, poorly documented scores
• Processing: Text cleaning, synonym mapping, stopword removal, stemming, and feature engineering

Modeling

• Approach: TF-IDF vectorization with Cosine Similarity for content-based filtering
• Baseline: Random selection (0% accuracy) vs. model (0.8 correct matches per profile)
• Optimization: Tested 19 TF-IDF configurations; best used 3000 features with 3-word minimum frequency threshold

Deployment

• Architecture: Offline training (pickle model) + online API (FastAPI/Flask) in Docker container
• Output: Top-N module recommendations with explainability (keyword matching reasons)
• Filters: Location, credits, level, start date applied post-recommendation

Ethical Considerations

• Transparency: Shows why modules are recommended
• Bias awareness: System reflects dataset limitations
• Privacy: GDPR-compliant (no personal data stored)
• Human oversight: Positioned as advisory tool, not automatic selection

Conclusion

The system successfully demonstrates AI-assisted module selection is feasible and valuable, provided data quality improves. Critical success factors include complete module descriptions, proper tagging, and maintaining human judgment in final decisions.

---

References

avans documenten

ID: M202512181408 Status: MOC Tags: Avans, projects

avans documenten

Voor avans moet ik veel inleveren, als groeps verband, maar ook prive. Maar ook moeten we veel bestanden maken voor ons zelf, om te zorgen dat het project voorspoedig verloopt. Aan de tags van het bestand kun je zien voor wel project en welk jaar het was.

---

Link to original