Analysis and modeling on COVID-19 related data.
-
Data Source Disease data in MongoDB Mobility data in MongoDB & Git
-
Data Processing Clean: Remove null, NA Query/FIlter: US States only, US county only Merge tables according to place and time: df.left(["State","date"])
-
Feature Selection Refer to references and work from others: census data, mobility data, current epidemics Lag Effect Analysis on mobility data a) Create lagged mobility sequence b) Filter to places with more than 50 cases in total c) statistical analysis/qualitative comparison
-
4E Model Building Equation: Derivative of Total cases in log scale ~ Lagged mobility data + days + population density + Aged Group >65 Estimation: Algorithm: Random Forest Modeling, Gradient Boost, MLR Evalutaion: MAE, MAPE, RMSE Training/Testing: Randomly 1/4 for testing, 3/4 for training Execution: A model ready to output prediction based on independent variables
-
Production Data Prediction for all level of administration, and for future