The goal of the project was to implement a fully functional web app in which a user can input a message related to a catastrophic disaster (e.g. earthquake, flood, fire,...) and would receive a list of potential categories which describe the content of the message. The classification output is based on a machine learning model trained with social media messages using natural language processing (NLP).

The project covers the fundamental steps of a machine learning project:

• Loading raw data
• Cleaning data
• Storing cleaned data
• Training & evaluating a machine learning model
• Storing the ML-model
• Building a web app to make use of the ML-model and providing a service to users

The project structure is made up of three main folders:

• app: Contains code related to the web app
• data: Contains the raw data files (.csv) and a script for data cleaning
• model: Contains a script to train and evaluate the machine learning model

To get started with the project run the following commands from the root directory:

1. python3 ./data/process_data.py ./data/disaster_messages.csv ./data/disaster_categories.csv ./data/DisasterResponse.db
2. python3 ./models/train_classifier.py ./data/DisasterResponse.db ./models/disaster_response_model.pkl
3. python3 ./app/run.py

Go to http://0.0.0.0:3001

Notes: You need to be within a Python virtual environment.

Special thanks to Figure Eight (acquired by Appen) for providing the curated dataset.