University of Twente Student Theses

Login

Understanding athletes preferences of sport activities through ranking based news feed recommendation

Perenzoni, Stefano (2022) Understanding athletes preferences of sport activities through ranking based news feed recommendation.

[img] PDF
2MB
Abstract:Product personalization is essential to provide users with personalized content and retain them. Recommender systems are algorithms which give suggestions on what the users may find more interesting. In this thesis, we want to implement a recommender system to personalize adidas Runtastic's news feed. Firstly, we analyse the historical data of Runtastic apps to understand the business context and needs. This thesis then proposes an approach to optimize diversity within ranking algorithms which does not rely on a diversification-aware loss function. Instead, we suggest Div-NDCG, a diversity-aware evaluation metric to be used in the model selection process to optimize both NDCG and the diversity metric by selecting the best model and dataset. Experiments are conducted on Runtastic historical data, from which we built multiple different datasets. The datasets are built through the combinations of different feature engineering and labelling techniques. Both heuristic-based and machine-learning models are then trained on tested on the datasets. By comparing their performances, we show that the diversity-aware evaluation metric better helps choose a model and dataset that optimize both metrics. Finally, we propose an architecture to implement the tested models on the currently deployed Runtastic apps to provide product personalization.
Item Type:Essay (Master)
Clients:
adidas Runtastic, Linz, Austria
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Programme:Computer Science MSc (60300)
Link to this item:https://purl.utwente.nl/essays/93669
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page