Back to Hui Li's Homepage

Dianping_Sigir14

This is the anonymized Dianping dataset used in our SIGIR'14 paper. It contains 11,352 users, 10,657 restaurants and 501,472 ratings from April 2003 to November 2013 in Shanghai, China. For the social friend network, there are a total of 280,041 claimed social relationships (directed edge). We cleaned those edges that point to users who do not have ratings in this dataset, while we used these edges in our experiments. Thus the edge number is smaller than we reported in the paper. Please see README file for details of data format. [download]

Dianping_RecSys15

This is part of the anonymized Dianping dataset used in our RecSys'15 paper. It contains 147,918 users, 11,123 restaurants and 2,149,675 ratings from April 2003 to November 2013 in Shanghai, China. For the social friend network, there are a total of 629,618 claimed social relationships (undirected edge). For privacy issue, we do not include the user information and restaurant attributes which can be used to identify a real person. Please see README file for details of data format. [download]

Dianping_Raw

For privacy issue, we do not publish the whole graph in raw data as well as the details of user information (e.g., birthday and taste tags) and restaurant attributes (e.g., whether it is 24 hour open and whether it provides parking space) which can be use to identify a real person. Besides, we also collected the review text in Chinese though we only publish rating scores. If you are interested in these information, fell free to contact us.

Citation

If you use Dianping dataset, please cite our papers [bib]:

[1] Hui Li, Dingming Wu, and Nikos Mamoulis. A revisit to social network-based recommender systems. In SIGIR, pages 1239--1242. ACM, 2014.
[2] Hui Li, Dingming Wu, Wenbin Tang, and Nikos Mamoulis. Overlapping community regularization for rating prediction in social recommender systems. In RecSys, pages 27--34. ACM, 2015.