Commit 03c0091e authored by Christian Marius Lillelund's avatar Christian Marius Lillelund
Browse files

Merge branch 'master' of https://gitlab.au.dk/cfp/air

parents 03b5c736 8df358aa
Pipeline #100649 passed with stage
in 4 minutes and 11 seconds
......@@ -3,6 +3,16 @@ AIR
AIR is an open-source machine learning project to improve rehabilitation.
AI-Rehabilitering (AIR) is one of seven government-funded signature projects, that aims to test and evaluate artificial intelligence in a municipal setting. More concretely, the AIR project focuses on the use of artificial intelligence as a tool to support municipal case workers, when they evaluate citizens, that are planned to participate in a rehabilitation course. Artificial intelligence can improve the level of service in the public sector, but there is little experience with such technology in the sector. The Danish Government (Regeringen), KL (Kommunernes Landsforening) and Danish Regions (Danske Regioner) has agreed that Danish municipalities should test artificial intelligence solutions to better understand how these can be utilized to improve the quality and capacity in the public sector in the future.
The goal of AIR is to develop a decision-support tool, that can provide case workers with an objective measurement of a citizen’s current eligibility for a certain rehabilitation course, their risk of failing during such a course and an estimate of how well the citizen is thought to do in the course. Current courses span 12 weeks and consist of physical exercises done several times a week by the citizen and evaluated at regular intermediate steps by the care assistant. Performance is measured in terms of improved self-sufficiency and reduced reliance on home help.
AIR is anchored in the Municipality of Aalborg and will pave the way for a solution based on models and statistics, that can support the individual professional assessment done by the case worker when it comes to choosing the right course of action of rehabilitation, including specific initiatives, improved utilization of assistive aids and fall prevention.
Read more at: https://projekter.au.dk/air/
<img src="https://digit.au.dk/fileadmin/_processed_/b/3/csm_AIR_illustration._Christian_Marius_Lillelund._6cd96175d4.png">
Project Organization
==============================
......
......@@ -7,7 +7,6 @@ Project Organization
==============================
├── LICENSE
├── Makefile <- Makefile with commands like `make data` or `make train`
├── README.md <- The top-level README for developers using this project.
├── data
│   ├── external <- Data from third party sources.
......@@ -41,23 +40,23 @@ Python == 3.8
Usage
==============================
To get up and running, clone the repository and copy the raw data to `air\data\raw\2020` or `air\data\raw\2019`, depending on your data version. Then go to do root project directory, install all modules and then install all requirements as follows:
To get up and running, clone the repository and copy the raw data to `air\data\raw\2021`. Then go to do the root project directory, install the src module and then all requirements as follows:
$ pip install -e .
$ pip install -r requirements.txt
Please note the project depend on scikit-survival, which requires Microsoft Visual C++ 14.0 or greater. Get it from: https://visualstudio.microsoft.com/visual-cpp-build-tools/
In the root project directory there is a client which can create datasets, make models and generate SHAP plots. It's a executable Python script. By default it will use the 2020 data version, encode the datasets as embeddings, not make visuzaliations of embeddings and not use the real ATS names, but their ISO id instead. To run the client:
In the root project directory there is a client which can create datasets. It's a executable Python script. By default it will use the 2021 data version and encode the categorial features of the datasets (the assitive aids data) as entity embeddings. To run the client:
$ python .\client.py -h
As an example, to run the client with the 2020 data version and with embeddings:
As an example, to run the client and encode the categorial features as one-hot-encoded columns:
$ python .\client.py --dataset-year '2020' --dataset-version 'emb'
$ python .\client.py --dataset-version 'ohe'
Contact
==============================
Please contact [Christian Marius Lillelund](mailto:cl@ece.au.dk) for questions regarding this repository.
Please contact [Christian Marius Lillelund](mailto:cl@ece.au.dk) or [Christian Fischer Pedersen](mailto:cfp@ece.au.dk) for questions regarding this repository.
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment