Application de l'analyse de données, des statistiques descriptives et de l'apprentissage automatique dans l'industrie aéronautique
This repository is my final work on the MVA course supervised by Jérôme Lacaille.
The objective of this work is to analyse 1000 flights. The analysis is contained in the file main.ipynb and is divided in 4 parts:
- loading and exploring the data
- extract the phases and especially the climbs
- cluster the climbs
- link with fuel consumption
To visualize and run the main notebook, you need to
git clone [email protected]:MatiasEtcheve/AERO.git
cd AERO
pip install -r requirements.txt
You also need to put the file Aircraft_01.h5 in the folder archive/.
Then you can simply run the notebook main.ipynb.
I have had a lot of problems when working on this project. The main problem was dealing with Plotly widgets. While Plotly offers a nice interactive views, the widgets don't work well / or at all when using Windows Subsytem for Linux WSL2.
This bug is not fixed, and I will open an issue on ipywidget.
Currently, it looks like the version ipywidgets==7.7.2 works the best. Below is the bug I obtained in most cases.
This repository is quite simple:
| File/folder | Usefulness |
|---|---|
archive/ |
Folder containing the dataset. The main notebook also write to this folder to save the climbs. |
tabata |
Tabata toolbox from https://github.com/jee51/tabata. Modified to add 2 features:* creating a dataset from a generator * seeing a phase with multiple values |
checks.py |
Contains all the basic checks to perform on a dataset |
main.ipynb |
Main work |
utils.py |
Contains functions to extract sequence in DataFrames |