COVID-19 Google Searches for Symptoms

Are Google searches for COVID-19 symptoms useful in predicting increases in COVID-19 patients? Exploratory data analysis and hypothesis testing using Google's COVID-19 Open Data database
Explore the docs »

Table of Contents

1. About The Project
   • Built With
   • Repository Structure
2. Data
3. Features
4. Contact

About The Project

COVID-19 has been a rapidly evolving pandemic. It is important for hospitals and governments to be able to predict the spread of COVID-19 and future pandemics locally, in order to prepare and allocate resources. This project provides tools to explore data from Google's COVID-19 Open Data as well as analysis and hypothesis testing.

Built With

Python 3.9:

Data modules

• pandas
• numpy

Plotting modules

• matplotlib
• seaborn
• scipy.stats

Misc. modules

• math
• warnings
• datetime
• itertools

Custom functions found here.

Repository structure

• img: Figure image files
• src: Custom Python functions and Jupyter Notebooks

A PowerPoint presentation summarizing the data analysis and results can be found here.

Data

This project explores 3 tables from the COVID-19 Open Data repo:

• Search Trends symptoms dataset
  • Dimensions: (1 425 194, ~450)
• Hospitalization records
  • Dimensions: (643 715, 11)
• Index: keys, codes, and names for countries and regions The search trends and hospital tables are joined using location keys and dates.

Features

• Load data with control over query size

• Generate plots for EDA to explore variables

  Example of hospitalization data versus time:

  

• Hypothesis testing of correlation between Google symptom searches and new hospitalizations

  • By location (country/region)
  • By symptom names
  • By date range
  • By time shift between searches and new hospitalizations

  Example of p-value heatmap:

  

Contact

Author: Christopher Shaffer

Email

Github