(Day 144) Using Graph Neural Networks to predict taxi passenger demand and origin/destination

Ivan Ivanov · May 24, 2024

Hello :) Today is Day 144!

A quick summary of today:

  • started reading papers on using graphs for taxi demand and origin/destination predictions

One of the papers my professor wrote is on Measuring taxi ridesharing effects and its spatiotemporal pattern in Seoul, Korea and because I am learning about GNNs from Stanford’s XCS224: ML with Graphs I remembered that one of the most popular use cases of graphs is on traffic prediction.

I read 2 papers:

I actually found the 2nd paper from the 1st paper because it was referenced.

There are these 4 models used as comparison vs the proposed GNN-based model

image

As for the models, the 1st paper kind of goes off from the 2nd one on the idea to separate a city into grids.

image

and construct an OD (Origin-Destination/Adjacency) matrix based on that with requests from one cell to another

image

Below is a pic from the 2nd paper on grids

image

Compared to the 4 models, the proposed GNN-based one performs the best on MAPE and MAE scores.

image

There are a few more papers that I have noted to read during the weekend:

Passenger Mobility Prediction via Representation Learning for Dynamic Directed and Weighted Graph

ST-LoRA: Low-rank Adaptation for Spatio-Temporal Forecasting - this one I want to read is because the 2 papers from above that I read - they use this grid cell for origin/destination demand and what if we can use LoRA on that matrix? Might idea might be totally off but I want to explore it.

A greedy approach for increased vehicle utilization in ridesharing networks

Fair and Efficient Ridesharing: A Dynamic Programming-based Relocation Approach

I hope I can think of something to research on these topics! Especially since one of professor Park’s interests are in transportation.

That is all for today!

See you tomorrow :)