This project builds an ETL pipeline to process disaster response messages and their categories. Then uses a ML pipeline to build a model to predict the category of messages, which can be entered by the user in a web UI. Check it out on https://disasterresponsepipeline.herokuapp.com/

The UI shows what categories the model predicted for the inputted message, by using 4 different ML models. This allows for better predictions and comparison between the models.

Here's a few examples

To run the project you need python3 and pip installed. The full list of dependencies can be found in the requirements.txt.

Clone the project

git clone https://github.com/luchev/disaster-response-pipeline

Enter project directory

cd disaster-response-pipeline

Set up a virtual environment

python3 -m venv venv

Activate the virtual invironment

source venv/bin/activate

Install requirements

pip install -r requirements.txt

The project comes with wrangled data and pre-trained models. If you want to use the pre-trained models you can just run the webserver:

python3 run.py

or

./runFlask.sh

Then go to http://0.0.0.0:3001/ to see the running project.

If you want to train your own models on custom dataset you will have to update the model path in app/run.py:34

• To run ETL pipeline that cleans data and stores in database python3 data/process_data.py data/disaster_messages.csv data/disaster_categories.csv <Database-Name>.db

• To run ML pipeline that trains classifier and saves python3 models/train_classifier.py <DatabasePath>.db <ModelPath>.pkl [ModelClassifier]

There are 4 classifiers you can use:

• RandomForest (default)
• AdaBoost
• KNeighbours
• DecisionTree

Process the sample data using the ETL (executed in the root of the project)

python3 data/process_data.py data/disaster_messages.csv data/disaster_categories.csv data/DisasterResponse.db

Train a new KNeighbours model (executed in the root of the project)

python3 models/train_classifier.py data/DisasterResponse.db models/new-kneighbours.pkl kneighbours

This project is developed as part of Udacity's Data Science Nanodegree Program

The data is provided by Figure Eight

This project is licensed under the terms of the MIT license.