lcombs/ML_COVID_PREDICTION

This project was completed by five masters students from New York University's Center for Data Science for our Machine Learning Class. It explores using machine learning algorithms to predict the health status of COVID patients in the Philippines.

Jupyter Notebook

ML_COVID_PREDICTION

Team Folder: https://drive.google.com/drive/u/2/folders/1zf3p0ORcRn2ngabcy2aFOnOvCt8DiOkR
Initial Data Source: https://news.abs-cbn.com/news/03/15/20/list-health-departments-list-of-confirmed-covid-19-cases?fbclid=IwAR1w1pcW5zoyL46ihgThrmQi4HUqcVibPXHBKEPptrxXGdT-cZAZpow2syo
Data Science Philippines: https://www.facebook.com/groups/datasciencephilippines
DSP Dataset Opensource: https://docs.google.com/spreadsheets/d/16g_PUxKYMC0XjeEKF6FPUBq2-pFgmTkHoj5lbVrGLhE/edit#gid=801084642
DOH Dataset PH: https://docs.google.com/spreadsheets/d/1BLbrvgjkBWxr9g73xX9DLOqmbmuYyKc-_b8jIxCX1uo/edit#gid=286423021
Philippine Statistics Agency for Population Data: https://psa.gov.ph/products-and-services/publications/philippine-statistical-yearbook
Philippine Open Data for Demographic Information: https://data.gov.ph/?q=search%2Ftype%2Fdataset&query=population&sort_by=changed&sort_order=DESC

Process:

Data Cleaning
- Geography Data:
  - Downloaded PH Case # Data
    - Used OpenRefine to clean 'Location' column
      - nearest neighbor, PPM
    - OHE on other columns
    - basic data cleaning (training spaces etc)
  - Downloaded DOH C Case Data
    - Used OpenRefine to clean 'Location' column and extract first element (city)
      - Clustering Methods: (https://github.com/OpenRefine/OpenRefine/wiki/Clustering)
        
        key collision method, fingerprint function
        
        key collision method, metaphone3
        
        nearest neighbor method, PPM distance function

Resources:

Geoencoding: