Using data, natural language processing, and machine learning to analyze and improve mental health conversations in the tech community.

• Mental Health in Tech App
• Background
• Team
• Notebooks
• Technologies Used
• Getting Started
• Database Schema
• API
• Starting the Flask App
• Using the pgAdmin app to view and query the database locally
• Deploying the app
• Issues

The results of this project are published at the following URL: https://mental-health-in-tech.herokuapp.com/.

This project looks at different maching learning classification models and compares their ability/accuracy to classify whether or not an employee in the tech community is willing to seek treatment for a mental health condition. The data used for this project comes from the OSMI (Open Sourcing Mental Illness) Mental Health in Tech Survey, which is an annual survey that captures employees' attitudes and opinions about how mental health is talked about and handled in the workplace in the tech industry. The target question from this survey that this project analyzes is "Have you ever sought treatment for a mental health disorder from a mental health professional?" Along with analyzing whether or not an employee would seek treatment, this project also uses natural language processing and sentiment analysis to build a model used to analyze mental health conversations employees have with their employers and then classify those conversations as either positive or negative.

Phil Stubbs

Jenna Nytes

Abby Lemon

Jennifer Cain

• AWS S3
• AWS RDS
• Postgres
• Python
• Google Colab
• Pandas
• PySpark
• Flask
• Swagger
• HTML/CSS
• Heroku
• TextBlob
• VADER

All of the ETL, data analysis, machine learning, natural language processing, and any other data processing task for this project was done in Python inside Google Colab notebooks. The notebooks are stored in Google Drive where everyone on the team can access them from the same location. They are also stored in this repository, but the notebooks in this repository might not be as up to date with the notebook files that are being worked on in Google Drive.

The following section will take you through the steps of setting up this project and getting it running locally on your computer.

If you don't want to set up this project locally and just want to see the deployed app, go to https://mental-health-in-tech.herokuapp.com/.

1. Clone the repository
2. Set up AWS and Google Drive
3. Load csv files from S3 bucket into Pandas dataframes to perform ETL
4. Create schema and tables

The first step is to clone the project repository to a local directory on your computer.

This project uses AWS to store the data files in a S3 bucket and uses Google Colab notebooks to extract, transform, load, and analyze that data.

To set up AWS and Google Drive for this project, run through the cells in the cloud_setup.ipynb notebook.

After setting up AWS and Google Drive, run through the cells in the cloud_etl.ipynb notebook to load the csv files from the S3 bucket into Pandas dataframes to extract, transform, and eventually load the data into a Postgres database.

Run through the cells in the schema.ipynb file to create the schema for the database tables and connect to the database to verify that the tables were created.

This project uses a SQL database hosted on AWS using AWS' Relational Database Service (RDS) to store the mental health survey results. For more information about the database structure used for this project, see Database Structure.

The data used for this project is stored in a Postgres database that is hosted on AWS using AWS' Relational Database Service (RDS).

To access the survey responses from the database, we built a simple API using Flask.

The API documentation is available in a swagger app. Start here for a basic reference on how to use the endpoints available to query the database for the data you need.

The API and frontend for this project are built using flask. To start the flask application locally:

1. Change directory to the application folder in this repository.
2. In the application folder, create a file called config.py that includes the following contents:

DB_USERNAME = 'username'
DB_PASSWORD = 'password'
DB_ENDPOINT = 'endpoint'

Replace username, password, and endpoint with their actual values.

3. Go up one directory to the root directory of this repository (mental_health_ML).

4. Run the following command to start the flask server on port 5000.

python application/app.py

Or, you can run the following shell script from the project root directory:

sh run.sh

4. Navigate to https://localhost:5000 in Chrome (or whatever browser you prefer) to view the app.

This project uses a Postgres database hosted on AWS RDS to store the data. If you have the pgAdmin app, you can view and query the database.

To connect to the database using pgAdmin:

1. Open the pgAdmin app.
2. Right click Servers.
3. Click Create > Server....
4. Enter a name for the server. For example, mental_health_tech_db.
5. Click the Connection tab.
6. In the Host name/Address field, enter the AWS RDS endpoint URL.
7. In the Username field, enter the database username.
8. In the Password field, enter the database password.
9. Check the Save Password? checkbox.
10. Click Save.

The app for this project is deployed to and hosted on Heroku. For more information on hosting with Heroku, see https://devcenter.heroku.com/. To deploy the app, you will need to have the Heroku CLI installed.

Note: This app use heroku-buildpack-python-nltk, which allows the model files used for natural language processing to be uploaded to the Heroku app. This is the same as Heroku's official Python buildpack but also installs any NLTK packages.

1. Download and Install the Heroku CLI.
2. If you haven't already, log in to your Heroku account and follow the prompts to create a new SSH public key.

heroku login

3. Clone the repository

Use Git to clone the repository to your local machine.

heroku git:clone -a mental-health-in-tech
cd mental-health-in-tech

4. Deploy your changes

git add .
git commit -m "changes"
git push heroku master --no-verify

If you find an issue while using the app or have a request, log the issue or request here. These issues will be addressed in a future code update.