salitr/disaster_response_pipeline

Disaster Response Pipeline

Building a web app that classifies disaster messages. The datasets used in bulding the Machine Learning Model are provided from Figure Eight.

Libraries

The following libraties should be imported; pandas, numpy, os, matplotlib json, multiple sklearn models, plotly, sys, re, pickle, warnings, sqlalchemy, NLTK, subprocess, termcolor, joblib, and flask

Motivation

The dataset contains real messages that were sent during disaster events. The repository include a web app where an emergency worker can input a new message and get classification results in several categories, and hence, the message will be sent to an appropriate disaster relief agency. The web app displaies a data visualizations of the overall training data.

File Descriptions

1. data Folder includes 4 files:

• disaster_messages.csv: Dataset includes all the messages and genres.
• disaster_categories.csv: Dataset includes all the categories.
• process_data.py: Code used to transforme and clean the data and then to create a SQLite database.
• DisasterResponse.db: SQLite database contains the transformed and cleaned data.

2. models Folder includes 2 files:

• train_classifier.py: Machine Learning Model to train and export a classifier as a pickle file.
• classifier.pkl: Final model as a pickle file.

3. app Folder includes 1 files:

• run.py: Flask file to run the web app.

Instructions

1. process_data.py:

• To process and run the ETL pipeline that transforms and cleans the data and the creates a SQLite database:

python process_data.py disaster_messages.csv disaster_categories.csv DisasterResponse.db

2. train_classifier.py:

• To train and run the ML pipeline that train and export a classifier as a pickle file:

python train_classifier.py ../data/DisasterResponse.db classifier.pkl

3. run.py:

• To run the web application that allows for adding new messages and then getting classification results in several categories:

python run.py

• Go to http://0.0.0.0:3001/

Licensing and Acknowledgement

• The datasets used were provided by Figure Eight.
• Code templates were provided by Udacity as a part of the Udacity Data Scientist Nanodegree.
• Plotly provides examples of codes for different types of plots of which some have been adapted, edited, and used as needed.