The Master’s Thesis presents the results achieved during the months of internship at Universidad Autónoma de Madrid, a Spanish public university, established in 1968. The main focus of this work is to provide a scientific analysis and a description of the implementation realized to develop a Recommender System using Graph Neural Networks. After analyzing the state of the art for traditional recommendation systems as well as the potential of Graph Neural Networks (GNN), such as GraphSAGE, the thesis focuses on the hybrid recommendation systems that employ GNN, which has piqued the interest of researchers because it promises to address some problems by augmenting existing information with graphs. This Master's Thesis describes an architecture for tackling the recommendation issue for the Point of Interest, which uses the Heterogeneous graph to characterize the data supplied to the system. Finally, a summary of the work done, and the future perspectives are given, with hints regarding a possible future research.
La tesi presenta i risultati raggiunti durante i mesi presso l'Universidad Autónoma de Madrid, un'università pubblica spagnola, fondata nel 1968. L'obiettivo principale di questo lavoro è fornire un'analisi scientifica e una descrizione dell'implementazione realizzata per sviluppare un Recommender System che utilizzasse reti neurali a grafo. Dopo aver analizzato lo stato dell'arte per i sistemi di raccomandazione tradizionali e il potenziale delle Graph Neural Network (GNN), come GraphSAGE che ha suscitato l'interesse dei ricercatori, la tesi si concentra sui sistemi di raccomandazione ibridi che utilizzassero GNN. Questa tesi di laurea magistrale descrive un'architettura per affrontare il problema delle raccomandazioni per punti di interesse, utilizzando grafi eterogenei per caratterizzare i dati forniti al sistema. Infine, viene fornita una sintesi del lavoro svolto, e le prospettive future, con cenni su una possibile ricerca futura.
RACCOMANDAZIONE DI LUOGHI DI INTERESSE ATTRAVERSO LA SPERIMENTAZIONE DI TECNICHE GNN SU GRAFI ETEROGENEI
DI PALMA, DARIO
2021/2022
Abstract
The Master’s Thesis presents the results achieved during the months of internship at Universidad Autónoma de Madrid, a Spanish public university, established in 1968. The main focus of this work is to provide a scientific analysis and a description of the implementation realized to develop a Recommender System using Graph Neural Networks. After analyzing the state of the art for traditional recommendation systems as well as the potential of Graph Neural Networks (GNN), such as GraphSAGE, the thesis focuses on the hybrid recommendation systems that employ GNN, which has piqued the interest of researchers because it promises to address some problems by augmenting existing information with graphs. This Master's Thesis describes an architecture for tackling the recommendation issue for the Point of Interest, which uses the Heterogeneous graph to characterize the data supplied to the system. Finally, a summary of the work done, and the future perspectives are given, with hints regarding a possible future research.I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14253/70544