Disaster Response Pipeline Project

Introduction

The aim of this project is to analyze disaster data from Figure 8 to build a model for an API that classifies disaster messages. The data set contains real messages that were sent during disaster events. A machine learning pipeline will be developed to categorize these events which can be sent as messages to an appropriate disaster relief agency. A web app will also be developed where an emergency worker can input a new message and get classification results in several categories. The web app will also display visualizations of the data.

Prerequisites

Following are the main libraries utilized:

• nltk
• sklearn
• sqlalchemy

ETL Pipeline

File data/process_data.py contains following functions:

• Load the datsets - messages.csv and categories.csv
• Merge the two datasets and derive the categories
• Clean the data
• Store the final cleaned dataste in a SQLite database

ML Pipeline

File models/train_classifier.py contains following functions:

• Load data from the SQLite database
• Split the data into train and test sets
• Build a text processing and machine learning pipeline
• Train and tunes a model using GridSearchCV
• Evaluate result on the test set
• Exports the final model as a pickle file

Web App

• Run the following commands in the project's root directory to set up your database and model.

• To run ETL pipeline that cleans data and stores in database -
  python data/process_data.py data/disaster_messages.csv data/disaster_categories.csv data/DisasterResponse.db

• To run ML pipeline that trains classifier and saves the model -
  python models/train_classifier.py data/DisasterResponse.db models/classifier.pkl

• Run the command in the app's directory to run your web app - python run.py

• Go to http://0.0.0.0:3001/