siwangs / disaster-resp-pipeline

Disaster Response Pipeline Project

Table of Contents

• Environment
• Instructions of Running the Programs
• Project Overview
• Structure of Projects
• Delivery

Environment

Beside standard Conda Library, follow library need to installed:

• argparse

Following packages needs to be downloaded from NLTK before first used:

• punkt
• stopwords
• wordnet

Instructions of Running the Programs:

1. Run the following commands in the project's root directory to set up your database and model. argparse is introduced instead of python default arg library from python.

   • To run ETL pipeline that cleans data and stores in database python data/process_data.py --msgcsv disaster_messages.csv --catcsv disaster_categories.csv --dbfilename DisasterResponse.db
   • To run ML pipeline that trains classifier and saves python models/train_classifier.py --sourcedb DisasterResponse.db --modeloutput classifier.pkl

2. Run the following command in the app's directory to run your web app. python run.py

3. Go to http://0.0.0.0:3001/

Project Overview

Data Engineering, NLP and Flask API were applied to build an APP/API which potentially help emergency department to pass related disaster agency faster.

Structure of Projects:

app:

templates(for Web API)
    go.html 
    master.html

Data:

disaster_categories.csv (disaster lable file)
disaster_messages.csv (disaster message)
DisasterResponse.db (output db file, NOT include in this git)
process_data.py (Main Program to clean and transform the source files for model to consume)

model:

classifier.pkl (Model output file, NOT including in the git)
train_classifier.py ( Main Program to read Source DB, train and save the model) 

Delivery:

A web API was created to show the summary of 36 disaster categories and functionality to categorized new disaster messages to its corresponding categories.

Examples and Screenshot

Categorize messages:

Distribution of genres:

Top and Bottom 10 Counts of Categories: