Disaster Response Pipeline

In this project using two datasets of distress mesages from three different channels, their response categories (36) and data, NLP and Machine learning pipelines we build an web application with the optimized classifier to predict in which response category the message is likelier to belong, reducing the potential reaction time of the responding organizations.

Installation

This project requires Python 3.x and the following Python libraries installed:

• NumPy
• Pandas
• Matplotlib
• Json
• Plotly
• Nltk
• Flask
• Sklearn
• Sqlalchemy
• Sys
• Re
• Pickle

Files Description

    disaster-response
      |-- app
            |-- templates
                    |-- go.html
                    |-- master.html
            |-- run.py
      |-- data
            |-- disaster_message.csv
            |-- disaster_categories.csv
            |-- messagesDB.db
            |-- process_data.py
      |-- images
            |-- header.jfif
      |-- models
            |-- classifier.pkl
            |-- train_classifier.py
      |-- notebooks
            |-- ETL Pipeline Preparation.ipynb
            |-- ML Pipeline Preparation.ipynb
      |-- README
      |-- LICENSE

ETL Pipeline

File data/process_data.py contains data cleaning pipeline that:

• Loads the messages and categories dataset
• Merges the two datasets
• Cleans the data
• Stores it in a SQLite database

ML Pipeline

File models/train_classifier.py contains machine learning pipeline that:

• Loads data from the SQLite database
• Splits the data into training and testing sets
• Builds a text processing and machine learning pipeline
• Trains and tunes a model using GridSearchCV
• Outputs result on the test set
• Exports the final and best model as a pickle file

Flask Web App

1. 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/messagesDB.db
   • To run ML pipeline that trains classifier and saves python models/train_classifier.py data/messagesDB.db models/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/ or http://localhost:3001/

Web Application Screenshots

Visual 1

Visual 2

Visual 3

Dataset

For the project we used two datasets,'messages.csv' and 'categories.csv',

Dataset messages.csv

Dataset categories.csv

Then we proceed to combine both sets of data to finally divide the category column into 36 individual columns that refer to the category to which each message belongs.

When dealing with several columns as a learning target, we would be talking about a multiclass learning model.

Model Performance

Tests have been carried out with RandomForest, AdaBoostClassifier and GradientBoostClassifier. From of all of them, we obtained the best score with AdaBoosClassifier. For the metrics, the "Macro" average was used since the dataset is unbalanced.

Total evaluation over all categories: Precision: 0.7763, Recall: 0.6473, FScore: 0.6793

License

Apache License

Version 2.0, January 2004

http://www.apache.org/licenses/

Conclusions

Since the dataset is unbalanced (Visual 2), we will proceed to use the 'macro average' for the model evaluation metrics. Hence the macro-average gives every class the same importance, and therefore better reflects how well the model performs.