Covid-19-Modeling

Goal

This project is to model and visualize Covid-19 cases based on county-level demographic and economic dataset

Data

• The Covid-19 data (from JHU-CSSEGIS)
• The geo-json data (from Plotly)
• The US county-level demographics and economy data (from US Census Bureau)

Exploratory Data Analysis

• Notebook or Notebook via nbviewer
• Visualize the spatial distribution of Covid-19 confirmed cases and deaths cumulative till 07-31-2020 in United States via Plotly.
• Developed data processing pipeline to process the Covid-19 data via Spark.
• Modeled the temporal evolution of Covid-19 confirmed cases via a logistic growth model.
• Rationalized the fitted model parameters and the impact on the current Covid-19 situation.
• Predict future Covid-19 cases ovetime for Orange County, CA.

Modeling of Covid-19 Cases with County-Level Demographic and Economic Dataset

• Notebook
• Leveraged the county-level demographics and economy data from US Census Bureau.
• Processed the data to join two datasets via a common key and performed data cleaning and missing value imputation.
• Visualized the correlation between features.
• Performed feature selection via LASSO Regression and reduced from 49 to 9 features.
• Developed an XGBoost model using selected features to model log10(confirmed_Cases) as a function of county-level demographics and economy information.
• Fine-tuned the model hyperparameters via Randomized Search thru Scikit-Learn and achieved an RMSE of 0.33 on hold-out set.
• Extracted the feature importance and obtained insight into how the local demographics and economy shape the Covid-19 pandemic spread.