Table of Contents

1. Description
2. Dependencies
3. Installation
4. File Descriptions
5. Results
6. Licensing, Authors, and Acknowledgements

Description

This project in Collaboration with Figure Eight is part of the fullfillment of the Udacity Data Science NanoDegree. The dataset here is provided by Figure Eight and contains pre-labelled tweet and messages from real-life disaster events. This project is aim is to build a Natural Language Processing (NLP) model to categorize messages according to predefined Categories.

The Project is seperated into 3 main parts

• ETL Pipeline for gathering the datasets and preparing them for the Maching learning Modeling step
• ML Pipeline for building a NLP model and exporting saving into a database
• Flask App for providing end user interactivity with the model and visualizations.

---

Dependencies

- Python 3.x.x+
- Machine Learning & ELT: Pandas, Numpy, Sciki-Learn
- Natural Language Processes: nltk
- Database: SQLalchemy (SQLite Database)
- Model Persistence: Pickle
- Web App and Data Visualization: Flask, Plotly

Installation

1. Clone the Repository 

    
2. Run ETL Pipeline
python data/process_data.py data/disaster_messages.csv data/disaster_categories.csv data/DisasteResponse.db

3. Run ML Pipeline

```python

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

```

4. Run Web App
 ```python
 
 cd app
 python run.py
 
 ```
 5. Access webapp on ```http://0.0.0.0:3001/``` on your browser

File Descriptions

There are 3 main parts

• data folder: Data
  • ETL Pipeline Preparation.ipynb: ETL pipeline notebook
  • process_data.py: Contain ETL pipeline python code for preparing data for ML pipeline
• models folder : Contains machine learning files
  • ML Pipeline Preparation.ipynb: ML pipeline notebook
  • train_classifier.py: Contain python code for running ML Pipelinne
• app: Web App and Visualizations run.py: Main Falsk app template folder: contain templates files

Results

1. A Model to run prediction on, after following installatin above the model is persisted to a pickle file in the data founder 'classifier.pkl'

2. After the was ran an average f1-score .94 was obtained, 94%

3. A web interface to test the model, inferfaces below

Fig1 - Home Screen

Fig2 - Enter Question

---

Fig3 - Predicted Results

Fig4 - Distributed Categories

Licensing, Authors, Acknowledgements

Credits to Figure Eight for the data.

Author

Fuh Austin

udacity-datascience-disaster-response-pipeline