CALCLOUD MACHINE LEARNING DASHBOARD

Interactive dashboard for analysis and evaluation of prediction models for CALCLOUD (Hubble Space Telescope's data reprocessing in the cloud package).

Run

DOCKER

docker pull alphasentaurii/calcloud-ml-dashboard:latest
docker run -d -p 8050:8050 alphasentaurii/calcloud-ml-dashboard:latest

View the running dashboard in your browser: http://0.0.0.0:8050/

Install (for development)

Using setup.py

git clone https://grit.stsci.edu/rkein/calcloud-ml-dashboard
cd calcloud-ml-dashboard
python setup.py install --user

Using virtual env

git clone https://grit.stsci.edu/rkein/calcloud-ml-dashboard
python virtualenv dash-venv
source dash-venv/bin/activate
cd calcloud-ml-dashboard
pip install -r requirements.txt

Once installed you can run the flask app locally

git clone https://grit.stsci.edu/rkein/calcloud-ml-dashboard
cd calcloud-ml-dashboard/calcloudML
python app.py

View the running dashboard in your browser: http://127.0.0.1:8050/

Model Performance Evaluation

Compare and evaluate model versions with roc-auc, precision-recall, keras history, and confusion matrix plots.

Accuracy vs Loss Barplots and Keras History (train vs test)

Receiver Operator Characteristic (Area Under the Curve)

Exploratory Data Analysis

Analyze data distributions, linearity and other characteristics.

Feature Scatterplots by Instrument

Feature Boxplots by Instrument

License

See LICENSE.rst for more information.