Disaster Response Classification

Table of Contents

• Installation
• Project Motivation
• File Descriptions
• Instructions
• Acknowledgments

Installation

The project used Python and HTML. The necessary libraries to run the code can be found in the requirements.txt file.

Project Motivation

Disaster organizations need to filter and pull out the relevant and most important messages from millions of messages following a disaster. Then the reflected problems are assigned to different professionals/organizations to take care of, such as water, medical and so on.

The project trained a classifier using the historical messages taken after disasters and the categories that the messages belong to. The classifier can predict the categories that a new message most likely belongs to for disaster response.

File Descriptions

The structure of the project and the file descriptions are as follows.

• app
  • template
    • master.html : main page of web app
    • go.html : classification result page of web app
  • run.py : Flask file that runs the app
  • img : include the images shown on the web page
• data
  • disaster_categories.csv : categories data to process
  • disaster_messages.csv : messages data to process
  • process_data.py : file that processes the data
  • DisasterResponse.db : database that stores clean data
• models
  • train_classifier.py : file that trains the classifier using cleaned data
• ETL Pipeline Preparation.ipynb : development process of process_data.py
• ML Pipeline Preparation.ipynb : development process of train_classifier.py
• requirements.txt : All required libraries to run the code for this project
• README.md

Instructions

1. To process the data and train the model, run the following commands in the project's root directory.
   • Run the ETL pipeline to process the data and store the cleaned data in the Sqlite database: python data/process_data.py data/disaster_messages.csv data/disaster_categories.csv data/DisasterResponse.db
   • Run the ML pipeline to train and save the classifier: python models/train_classifier.py data/DisasterResponse.db models/classifier.pkl
2. To run the web app, run the following command in the app directory: python run.py

Screenshot of the Web App

1. Before typing in anything, the main page shows three visualizations that provides an overview of the training dataset.

2. To classify a new message, type in search box and click the "Classify Message" button.

The predicted categories are highlighted.

Acknowledgments

The disaster data are from Appen. The skeleton code for this web app is provided by Udacity Data Scientist Nanodegree.