Skip to content
AutoML project for ML Products class
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
R
automl
sample_data
.gitignore
README.md
requirements.txt

README.md

mlp-automl R version

AutoML project for ML Products for Data Science. This tool allows users to upload datasets and with one click, perform

  • Descriptive Statistical analyses
  • Train different models and
  • Evaluate the trained models using AUC

Instructions for Adam Kelleher

All the code for the R project is under R directory

Project Structure

We used R and Shiny for this project. All files are located under the R directory. The shiny.R file is our app entry point as well as the dashboard. We have separated model training, prediction and evaluation into different services. As well as computing descriptive statistics and preprocessing the data.

To run the application, simply clone the repo and launch shiny.R

mlp-automl python version

Rest API that allows users to upload datasets and perform some statistical analyses, plot a pairplot and evaluate (AUC) a machine learning model.

Python version

For best results use Python 3.x

Dependencies

Virtual environment

It is best to run this project with on a venv

Install virtualenv

pip install virtualenv

Create a virtual environment for project in a separate dir

virtualenv -p python3 mlp-automl-env

Activate virtualenv. Navigate to venv dir and run

source <path-to-venv>/bin/activate

Later, to deactivate venv run

deactivate

Redis Message Broker

Install the message broker Redis following instructions here:

https://redis.io/download

Package dependencies

From project root dir, install dependencies with

pip install -r requirements.txt

REST Call to the web service

  • Input data files can be found in the dir: sample_data
  • Make sure the package `gunicorn' is installed
  • Navigate to root directory
  • Open three terminal tabs and activate venv in each
  • Start the webservice with gunicorn automl:app one terminal
  • Test that the service works by navigting to localhost:8000/
  • Start redis server with src/redis-server in 2nd terminal
  • Start celery worker in last terminal tab with celery worker -A automl.celery --loglevel=inf
  • Create a new job with curl -XPOST -F 'data=@<path-to-data-file>/iris.data.txt' -F 'target=@<path-to-dataset-dependent-variable-file>/iris.target.txt' localhost:8000/job/ The app returns a jobid which will be used to download the results
  • Download job results from browser with localhost:8000/job/<jobid>
You can’t perform that action at this time.