DisasterResponse

Created a pipeline to filter and classify messages sent during disasters

Table of Contents

1. Background
2. Run Your own Disaster Response Classifier!
   1. Packages needed
   2. Installing
   3. Running Program

Background

When a disaster strikes, there are a flood of messages that disaster response teams have to respond to. The goal of this project is to creat a classifier that is able to classifier these methods into different categories in order to help the response teams

In order to do this I first created an ETL pipeline that extracts the data, cleans it, and then uploads it to a database. Then I created a machine learning pipeline that loads the data from the database, and then trains a multioutputclassifier model on the text from the database. From here I was able to manipulate parameters to get my final accuracy score up to about 93%.

Run Your own Disaster Response Classifier!

Packages needed

• Python (3.5)
• pandas
• numpy
• nltk
• sciki-Learn
• SQLalchemy
• Flask
• Plotly

Installing

Simply download these files directly or clone this repository using git.

Running Program:

1. The following commands will execute the program and create both pipelines that we need to run the web app. Make sure you are in the projects home directory when running

   • ETL Pipeline (cleans and stores data) python data/process_data.py data/disaster_messages.csv data/disaster_categories.csv data/DisasterResponse.db
   • ML pipeline (trains classifier) python models/train_classifier.py data/DisasterResponse.db models/classifier.pkl

2. Run the following command from the "app" directory to get your web app running. python run.py

3. Go to http://0.0.0.0:3001/ to see your running web app! From here you'll be able to type in disaster messages of your own and see how they get classified.