This paper investigates how a latent space model with link prediction as the network evaluator performs if implemented as a social recommendation system for restaurants on the Yelp platform. The model is investigated with both users and restaurants as the nodes of the networks. The model used is an optimized latent space model built for large scale networks by Nicolai Frost Jacobsen (2018)to be utilized in his master's thesis, Large scale latent variable modelling for link prediction in complex networks. Latent space models have had much success in other areas such as friend recommendations, movie recommendations and even proteins network. The latent space model and link prediction model performed well on all the networks investigated, though the conditions they performed well under indicated that the latent space model might not be the right fit to recommend restaurants within a large-scale network across states. The latent space became an expression of physical distance as the nodes within each state clustered together. The computational costs are too high compared to the value it creates in the form of social recommendations.
|Educations||MSc in Business Administration and Information Systems, (Graduate Programme) Final Thesis|
|Number of pages||59|
|Supervisors||Thomas W. Frick|