waterwheel31/disaster-response

Machine learning pipeline to analyze texts to classify. This is a project of Udacity Data Scientist Nanodegree

Disaster Response Machine Learning Pipeline

What is this?

This is a simple web app to read disaster response messages and judge what kind of messages they are.

How the algorithm is?

• Using Logistic Regression to judge whether each message is categorized to each of 36 categories
• There are 36 models working parallel, and a message can be categorized in multiple categories
• As for the preprocessing, Tf-idf is applied

Instructions for running:

1. Run the following commands in deploy directory to set up the 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 sqlite:///data/database.db
   • To run ML pipeline that trains classifier and saves python models/train_classifier.py sqlite:///data/database.db models/finalized_model.pkl

2. Run the following command in deploy/app directory to run the web app. python run.py

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

4. Input a sentence on the input box and click the button below, then the analysis results will appear.