https://github.com/pyladiesams/conformal-prediction-jan2024

An introduction to conformal prediction
https://github.com/pyladiesams/conformal-prediction-jan2024

conformal-prediction forecasting python

Last synced: about 1 month ago
JSON representation

An introduction to conformal prediction

• Host: GitHub
• URL: https://github.com/pyladiesams/conformal-prediction-jan2024
• Owner: pyladiesams
• License: mit
• Created: 2023-11-19T16:39:15.000Z (about 1 year ago)
• Default Branch: master
• Last Pushed: 2024-01-31T14:58:24.000Z (11 months ago)
• Last Synced: 2024-08-03T17:12:04.879Z (5 months ago)
• Topics: conformal-prediction, forecasting, python
• Language: Jupyter Notebook
• Homepage:
• Size: 3.22 MB
• Stars: 16
• Watchers: 3
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-conformal-prediction - conformal-prediction-jan2024 - PyLadies Amsterdam

README

        
# An introduction to conformal prediction

### Slides of presentation: [Conformal prediction](workshop/2024-01-PyLadies-Amsterdam-meet-up-Conformal-prediction.pdf)

### [Video](https://www.youtube.com/watch?v=--WcrDRtrYk&t=1s) of the presentation: Leveraging conformal prediction for calibrated probabilistic time series forecasts - Inge van den Ende, PyData Eindhoven 2023

### Workshop summary in a [blog post](https://medium.com/@icvandenende/leveraging-conformal-prediction-in-python-to-accelerate-the-renewable-energy-transition-09b5c855f69d)

## Workshop description

During the workshop we will apply conformal prediction:

1. Add a prediction interval to a point forecast 

2. Calibrate a probabilistic forecast (conformalized quantile regression)

Open the notebooks in the workshop folder to make the assignment. You can find the full code in the 

notebooks in the solutions folder. Both notebooks show how implement that step in the `crepes` 

package, the `MAPIE` package and manually by yourself. 

## Requirements

Python >= 3.10 and experience with making a regression forecast. In the workshop we take a LightGBM

model as an example to make a forecast, but this model is not explained in the workshop.

## Usage

* Clone the repository

* Install requirements with `pip install -r requirements.txt`

* Start jupyter lab and navigate to the workshop folder

* Open the first workshop notebook

## Video record

Re-watch [this YouTube stream](https://youtube.com/live/QFtdTyIWrz8)

## Credits

This workshop was set up by @pyladiesams and @ingevandenende.