OpenSiteRec

Usage

Each folder of Chicago, NYC, Singapore and Tokyo contains the OpenSiteRec data of the city. In each folder, the structure is as follows:

• City/
  • City_KG_plus.pkl/.csv
    • the main file that each instance denote a POI with corresponding information
  • City_Region.geojson
    • the geojson file that collect all the regions as Polygons in the city
  • City_BA.geojson
    • the geojson file that collect all the business areas as Polygons in the city
  • City_brands.pkl/.csv
    • the relation file that represent the relations between brands
  • City_regions.pkl/.csv
    • the relation file that represent the relations between regions

Please use the python package Pandas and GeoPandas to load the file:

import pandas as pd
import geopandas as gpd

df = pd.read_pickle('xxx.pkl')
gdf = gpd.read_file('xxx.geojson')

The folder baseline contains the implementation code of the baselines.

We also provide a demo data of Singapore using .csv format in folder demo.

Benchmarking Code

To facilitate the usage, we provide the code of baseline model training, where you can find the code to load the dataset in baseline/data_utils.py and to specify your training schedule in baseline/main.py.

However, this data loader merely organize the data in a simple and vanilla way. It is encouraged to play with the data according to your own idea.