The Berkeley SUMSarizer
The SUMSarizer reduces raw SUMS data into a log of "cooking events" using machine learning. Upload your CSV files, train the algorithm by manually annotating a few data sets (with a nice UI), and the SUMSarizer will do the rest.
Guaranteed to save you a month of data wrangling!
Check out the sample development environment for getting started on SUMSARIZER development.
SUMSarizer is a Python 2.7 app.
Install its requirements:
pip install -r requirements.txt
R should also be installed with the following packages:
plyr
devtools
pspline
caTools
glmnet
devtools::install_github('jeremyrcoyle/origami')
The current config expects a Postgres database at the following URI: postgresql://sums:sums@localhost/sums
The following environment variables should be set:
APP_SETTINGS="config.DevelopmentConfig"
python run.py
There are a few different background workers to spin up for various tasks:
python worker.py
python -m tasks.ml_worker
To initialize the DB schema run:
python manage.py db upgrade
If you change a model file in models.py, run:
python manage.py db migrate
to generate the migration file. Inspect the migration file in migrations/versions/...py If it looks good, run it with:
python manage.py db upgrade