A cry for ST-GCN help
I will be honest. I don't have much time (around a week) and I need to create ST-GCN neural network for my classes, but I know next to nothing about it and while using AI create "some" model, by looking at results, I see that it has nothing do to with ST-GCN and propably have data leakage.
So to what I need to create is a STGCN model that will predict bike usage in public bike sharing system (like Citi Bike in NYC) by using agregated data from 1 hour windows (last 5 hours should be used to predict result from next hour)
Currently I have data:
- List of station
- Usage over 3 month time (I am using data gathered from my city, it doesn't have public data, can only gather current state via API)
- Number of features for each station (like number of available bikes, temperature, rain, wind, distance to metro station etc.)
- Adjacency matrix for stations
Data is agregated for 1 hour windows where bike availability is an average from that hour.
Data is segregated in 3 sets, training, validated and test that are separate.
So from what I know, it looks like I have all the data I need, but I have honestly next to 0 idea how to build ST-GCN. What I am asking is something that will explain how to build it in python, because I don't know what to do or even when to start with it.