search

anderzzz / london_bike_forecast

Graph convolutional neural network for forecasting traffic in the London bike-share system, where the graph convolutions pass spatial information between stations, and one-dimensional convolutions pass information from past traffic.
5 stars 5 forks source link

station consts file #1

Open cxshine opened 3 years ago

cxshine commented 3 years ago

Hi, I noticed that you have done a great work for predicting bike sharing demand. I wanted to reproduce your work, but when I looked at raw_data.py file, I don't know how many files I need to download from database, and I also can't get the station const file, can u help me ? Thanks.

anderzzz commented 3 years ago

Hello.

Some months after I did the analysis in this repo, I uploaded most of the data (somewhat cleaned up) to the Kaggle dataset server. You may have to slightly massage the files to fit into the graph convolutional networks (like binning the data along the time axis, maybe discard some low-usage stations), still some of the messier parts of the raw data from the City of London website have been sorted out: https://www.kaggle.com/ajohrn/bikeshare-usage-in-london-and-taipei-network

You can also see the file london_stations.csv, which I believe is the specific station constant file you ask about. It can be found on City of London data website, but yes it is a bit hidden I recall. Unfortunately I did not find the equivalent data for the Taipei network, hence the absence of a taipei_stations.csv file.

Hope this helps.