Full description of the project please see final.md

There are two program files in this repo:

• model_training.ipynb This notebook prepares training data, trains the delay prediction model, and pre-compute the delay inference results on the test data, for each area in each year. We have saved the inference results for both Lausanne and Zurich on 2024, if you just want to test our application on a travel date between 2023-10-30 and 2024-11-05 in either Lausanne or Zurich, you don't need to execute this notebook and can directly run robust_journey.py; otherwise, you may train the model and make delay inference results first using this notebook with your customized uuids and time, which takes 10-20 min to run. Further instruction see the beginning of the notebook.

  UUIDs we have pre-computed on are:

  {

  ​ "Lausanne": ["a7a21b73-6ffe-4fbf-a635-6e2b961f3072", "e168fd57-f57a-4075-a350-0dcfbb55147f"],

  ​ "Zurich": ["69706755-dbb7-45d4-a424-5b9ec36f4d33"]

  }, and both are pre-computed from 2023-10-30 to 2024-11-05

• robust_journey.py This notebook contains all the necessary functions for journey planning and includes the interface that takes user inputs and returns a list of suggested journeys (as is required in final.md). The user interface is in the Journey Planning section and has both programmatic and visualized formats. If you'are testing on an area other than Lausanne or Zurich, or on a travel date beyond of 2023-10-30 and 2024-11-05, make sure you have pre-computed the delay inference results with model_training.ipynb, and updated the area_uuid_dict accordingly. The query speed varies from 1 min - 10 min.

Link to video presentation: https://youtu.be/Tl6amTjTZBs