Data Statistics
OpenSiteRec is an open dataset for Site Recommendation.
It consists data of four international metropolises: Chicago, New York City, Singapore and Tokyo.
There are five types of entities: Brand, Category, POI, Business Area, Region.
The extensive relations provide sufficient commercial and geographical information.
Statistics
Chicago, New York City, Singapore and Tokyo
Data statistics of OpenSiteRec.
Annotation
Chicago, New York City, Singapore and Tokyo
Annotations of OpenSiteRec.
Brand Category
Chicago, New York City, Singapore and Tokyo
The distributions of brand category are different across cities.
Brand Site Number
Chicago, New York City, Singapore and Tokyo
The long-tail problem is severe in OpenSiteRec.
Chicago
Population: 2.7 million | Area: 590 km2
Chicago is the most populous city in the U.S. state of Illinois and the third most populous in the United States.
New York City
Population: 8.8 million | Area: 778 km2
New York City or NYC, is the most populous city in the United States.
Singapore
Population: 5.6 million | Area: 733 km2
Singapore is a sovereign island country and city-state in maritime Southeast Asia.
Tokyo
Population: 9.7 million | Area: 620 km2
Tokyo, officially the Tokyo Metropolis, is the capital and most populous city of Japan.
Chicago
Population: 2.7 million | Area: 590 km2
Chicago is the most populous city in the U.S. state of Illinois and the third most populous in the United States.
New York City
Population: 8.8 million | Area: 778 km2
New York City or NYC, is the most populous city in the United States.
Singapore
Population: 5.6 million | Area: 733 km2
Singapore is a sovereign island country and city-state in maritime Southeast Asia.
Tokyo
Population: 9.7 million | Area: 620 km2
Tokyo, officially the Tokyo Metropolis, is the capital and most populous city of Japan.
As shown in the figure, the distributions of brand category are different across cities.
Such a phenomenon indicates the potential life-style preferences of people in each city.
As shown in the figure, the data of all four metropolises suffer from severe long-tail problem on the brands.
Generally speaking, the top 10% brands occupy over 50% of POIs.
Since Tokyo government has not released the official planning of business area, we can not visualize the business areas.
Benchmarking experimental results of 16 representative baselines on OpenSiteRec.
For Machine Learning methods, we have LR, GBDT, SVR and RankNet.
For Collaborative Filtering methods, we have MF-BPR, NeuMF, FISM and NAIS.
For CTR Prediction methods, we have DNN, Wide&Deep, DeepFM, xDeepFM.
For Graph-based methods, we have GC-MC, GraphRec, NGCF and LightGCN.
Please find the implementation codes at Github Repo OpenSiteRec.
