Analytics Value Chain Project

Make recommendations for items a user may want to buy based on the items they are adding to their shopping cart to anticipate items they may want to add, increase app engagement, and ultimately increase order size.

Increase app engagement, identify desired items, and ultimately increase order size

Increase order size by using recommendations from the user?s current shopping list to suggest missing items or items that the user may want to try, and make adding such items to the shopping list highly accessible to the user

User feedback and assessment logical order product sets

• Flask
• FlaskSQLAlchemy -
  • A flask plugin for SQLAlchemy. SQLAlchemy is an Object Relational Mapper (ORM), which means it allows interaction with relational data models using object oriented approaches, like those typically used in python.
  • This project uses SQLAlchemy to create, read from, and write to relational databases.
  • SQLAlchemy's flexibility will allow for a smooth transition from using a local database to using something like Amazon RDS. All that needs to change is a configuration in the app code (see this blog for more on transition to RDS)
• Surprise
  • Surprise is a Python scikit building and analyzing recommender systems
  • Surprise is used to build the a collaborative filtering recommender system for the app

1. Clone repository

2. Create virtual environment for new app

   virtualenv -p python3 Instacart

3. Activate environment

   source activate Instacart

4. Install required packages

   pip install -r requirements.txt

5. Download InstaCart csv files from Kaggle and save to analyze/data folder

6. Set up instacart.env file with the following structure to connect to a database instance:

   
   export DATABASE_URL= XXX
   
   export DATABASE=XXX
   export USERNAME=XXX
   export PORT=XXX
   export PASSWORD=XXX
   export HOST=XXX`
   
   export SECRET_KEY=XXX`` 
   
   

7. Set your environment

   source instacart.env

8. Define database

   python create_db.py

9. Create features, keys, and model by running make all from the analyze/ directory

10. Run tests on code

    cd analyze/tests
    pytest
    

11. Run the app by running python application.py from the root directory

You should be able to go to the IP address that it responds with and see your web app.

To see what it should look like, visit this link

Pivotal Tracker

Project Developer: Sarah Greenwood, Project Manager: Logan Wilson, QA: Jill Fan