By Molly Srour Lead QA Tova Simonson
Vision: To enable beer-novices and craft beer enthusiasts alike to discover new beers that fit their interests. The craft beer industry has been exploding over the last 10 years, and it could be considered daunting to try to navigate the huge variety in beer styles to find one that is appealing. Users will explore new beers and gain a greater appreciation of the options of beer available. After visiting this app, users will be more likely to buy more beer and frequent a wider range of breweries.
Mission: The app will be easy to navigate and will allow users to explore 10 diverse categories of beer and give targeted recommendations of specific beers in each category. Users will choose a category, name beers they find appealing, and be compared to a database of over 1.5 million total reviews of beer to find their perfect brew.
Success criteria:
Machine Learning Criteria: The main metric which will be used to evaluate app performance is accuracy. For first time users, an accuracy of 0.4 (2 of the 5 recommended beers) will be achieved. For second time users, we hope to achieve 0.6 accuracy and for third time users and above we will achieve 0.8 accuracy. We define a second time user as a user who has come back to the app and rated beers that were previously recommended to them. A user who visits the app once and gets recommendations for multiple categories of beer is still considered a first time user.
Business Success Criteria: The app will be considered successful from a business standpoint if 80% of first time users return to the app to log their ratings of their recommended beers and gain new recommendations. Additionally, the app will be successful if 10,000 new users use the app in the first 3 months. Users will need to specify a username to be added to the database, and this username will be used to track user return and userbase size.
├── README.md <- You are here
├── data <- Folder that contains data used or generated.
├── deliverables <- Folder that contains deliverables for AVC project
├── models <- Trained model objects (TMOs), model predictions, and/or model summaries
├── src <- Source data for the project
│ ├── config.yml <- Configuration file for model pipeline
| ├── acquire_data.py <- Script for getting data from my S3 bucket to your S3 bucket and finally to your local path
| ├── clean_data.py <- Script for cleaning data
| ├── train_model.py <- Script for training model on 450 different test users and creating predictions, top 10 beers for each beer type, and test user combinations
| ├── score_model.py <- Script for scoring model performance on a subset of the data
│ ├── configure_db.py <- Script for creating db in RDS or Sqlite and adding rows from the three tables
│ ├── getdata_s3.py <- Script for getting source data from my S3 bucket and landing it in your S3 bucket
│ ├── test_clean_data.py <- Unit tests of the clean_data script
│ ├── test_train_model.py <- Unit tests of the train_model script
│ ├── test_score_model.py <- Unit tests of the score_model script
├── static <- Static CSS folder
│ ├── basic.css <- CSS for the app
├── templates <- HTML templates for the app
│ ├── first.html <- HTML Landing page for the app, user chooses type
│ ├── recommender.html <- HTML page where user inputs beers they like from top ten list
│ ├── results.html <- HTML page where user receives results
├── application.py <- Script for running the Flask application
├── flask_config.py <- Configurations for Flask app
├── Makefile <- File to link together and create necessary files, models, etc
├── requirements.txt <- Python package dependencies
This project structure was partially influenced by the Cookiecutter Data Science project.
curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py
python get-pip.py
Clone this repo:
git clone https://github.com/mollysrour/beerapp/
Then enter the local folder
cd beerapp
The requirements.txt file contains the packages required to run the model code. An environment can be set up in two ways.
pip install virtualenv
virtualenv beerapp
source beerapp/bin/activate
conda create -n beerapp python=3.7
conda activate beerapp
Then you must install the required python packages, which are stored in the requirements.txt file.
However, due to an issue with the surprise package, you will need to install numpy first separately.
Documented here: NicolasHug/Surprise#188
Please run this first:
pip install numpyThen run this command:
pip install -r requirements.txt-
Set your AWS environment variables by running the following commands, entering your AWS key and secret key, all must be in quotes:
export AWS_KEY_ID="" export AWS_ACCESS_KEY=""
-
If you are using RDS, set your RDS MYSQL environment variables by running the following commands, entering your relevant RDS info, all must be in quotes:
export MYSQL_USER="" export MYSQL_PASSWORD="" export MYSQL_HOST="" export MYSQL_PORT=""
Additionally, if you are using RDS, also run this command:
export localorRDS="RDS"If you are NOT using RDS and are configuring the app locally, run the following command:
export localorRDS="local"Change the configurations in src/config.yml to your desired configurations. Mandatory changes include:
Within the acquire_data section of the config, change AWS_BUCKET to the name of your AWS bucket, change AWS_FILE_PATH to the file path within your bucket that you would like the raw data to be landed, and change localfilepath to the filepath in your local folder you would like the raw data to be landed. Within the configure_db section of the config, if you are configuring the app locally, you must change SQLITELOCALENGINE to the appropriate path for the SQLite database. This MUST be of the format 'sqlite:///filepath/databasename.db' . For example, if you are in your beerapp folder and wish the db to be stored in the data folder as beers.db, then the appropriate format of the SQLITELOCALENGINE is 'sqlite:///data/beers.db'. If you are using RDS, you must configure MYSQL_USER, MYSQL_PASSWORD, MYSQL_HOST, MYSQL_PORT, MYSQL_DB, and MYSQL_SQLTYPE to your appropriate RDS settings.
Go into the makefile and change the data paths appropriately to where you would like the data to be stored.
Run make all
Make all runs the acquire_data, clean_data, train_model, score_model, AND configure_db scripts. Only run this command once, because if you run it multiple times, it will duplicate the rows in the database. Make sure all your sqlite/rds configurations are properly exported and entered in the YAML before running this command.
flask_config.py holds the configurations for the Flask app. If you are configuring the app locally, you must change the SQLALCHEMY_DATABASE_URI to the same path that you specified in SQLITELOCALENGINE in the config for your database. If you are configuring the app using RDS, you should not need to change anything in flask_config.py since the os environment variables are used. HOWEVER, if your database is not called msia423 and your connection type is not mysql+pymysql, you must change those to their appropriate settings within flask_config.py.
Run python application.py
Run py.test to unit test this app. The test scripts are found in the src folder.
Functions that merely compiled other unit tested functions were not tested.