Home->Download

Datasets


gold
GOLD (Geo-Tagged Large Scale Dataset)

GOLD contains more than 22K Flickr Crawled images together with their Geo-tags and it covers 80 different places of interest mined from a Flickr dataset which contains 3.3 million images.

 

 

goldGOLDEN (Extended Geo-Tagged Large Scale Dataset)

GOLDEN is a larger scale geo-tagged image set with 5.2 M images from 1447 places of interest all over the world. GOLDEN is also crawled from Flickr. The selection of 1447 places is referred to list of places of interest all over the world from WIKI.com. There is no content overlapping for GOLD with GOLDEN. That is to say the GPS locations in GOLD are not appeared in GOLDEN, and vice is versa.


goldYelp Dataset: for Social Media Recommendation

The shared Yelp Dataset can be utilized to evaluation performance social media recommendation. This dataset consists of 10555 real users and their relationship with their friends. The shared Yelp data contains 8 categories, which we crawled form Yelp. Most of the users of these categories are from NYC. In each category, there are ratings, users, items, training and test sets file. The dataset can be utilized to evaluate social media recommendation performances objectively.


goldService Quality Evaluation Dataset: for Service Quality Evaluation

Service Quality Evaluation dataset consists of three categories: Restaurants, Nightlife, and Douban Movie. The data of first two categories are crawled from Yelp.com website. The Douban Movie data are crawled from movie.douban.com. The target is to predict the overall rating of services with few user reviews.




goldDouban Movie Dataset: for Social Media Recommendation

The shared Douban Movie Dataset can be utilized to evaluation performance social media recommendation. This dataset consists of more than 900K ratings from 2965 real users who rated a total of 39,695 movies. In the shared Douban Movie dataset, there are ratings, users, and items files, which are downloaded form www.douban.com website. The dataset can be utilized to evaluate social media recommendation performances objectively.









 
ҳģҳģ