Abstract
Recommender systems are built using traditional techniques such as collaborative filtering and content-based, and advance techniques like social network analysis. This study aims to explore and help people experience the top where-to-eat places by creating a recommender system that could suggests new restaurants to try-out.
A restaurant-network based recommendation technique was developed to generate restaurant recommendations for the users of Yelp online platform. The network created sets the restaurants as the network nodes and the number of Yelp users who wrote reviews on 2 restaurants as the network link between these restaurant nodes. The network properties and centrality measures were used as alternative metrics to identify and rank restaurant recommendations. Restaurant-networks were created separately per city from different states available in the dataset.
The states, cities and restaurants were identified by performing geospatial analysis on the user reviews combined with business location information. The developed system has demonstrated that eigenvector centrality measure recommends highly rated restaurants while other measures recommend based on cuisine or food type of previously visited restaurants.
