An Evaluation of Yelp Dataset

Yelp is one of the largest online searching and reviewing systems for kinds of businesses, including restaurants, shopping, home services et al. Analyzing the real world data from Yelp is valuable in acquiring the interests of users, which helps to improve the design of the next generation system. This paper targets the evaluation of Yelp dataset, which is provided in the Yelp data challenge. A bunch of interesting results are found. For instance, to reach any one in the Yelp social network, one only needs 4.5 hops on average, which verifies the classical six degree separation theory; Elite user mechanism is especially effective in maintaining the healthy of the whole network; Users who write less than 100 business reviews dominate. Those insights are expected to be considered by Yelp to make intelligent business decisions in the future.