Graph Learning Based Fashion Recommendations

Results of this project are explained in the final report. and a short demo video was made by the students, see video below:


  • Sponsored by: inovex GmbH
  • Project Leader: Dr. Ricardo Acevedo Cabra
  • Scientific Lead: Dr. Lea Petters, M.Sc. Frauke Beccard and Dr. André Ebert
  • TUM Co-Mentor: Cristina Cipriani
  • Term: Winter semester 2022


As generating relevant recommendations for customers is a key determinant of customer engagement and satisfaction as well as business success, the aim of this project is to implement a prototypical recommender service for a large fashion dataset. Filtering a very large item space for items that facilitate the customer’s buying decision is a complex task. In order to be able to capture the multi-dimensional relationships between items, users and attributes, we want to apply graph learning approaches, expecting that they will outperform traditional recommendation approaches in terms of quality of resulting recommendations.
After a thorough exploratory data analysis, it will be necessary to design and implement a graph database as well as a graph learning based recommender system (GLRS) for the fashion dataset. The recommender system will then be deployed, monitored and modified through multiple iterations.

Students should be familiar with python programming and git versioning and should have knowledge in the field of recommendation algorithms. Experience with Graph Learning
approaches and Graph Databases would be helpful.

Accepted students to this project should attend (unless they have proven knowledge) online workshops at the LRZ from 10.10. - 14.10.22. More information will be provided to students accepted to this project.