Predicting Average Adjusted Gross Income (AGI)

Goal: Develop linear regression model (Spark ML) on NYU DUMBO to Predict Average Adjusted Gross Income (AGI)

Input

Counts of businesses by NAICS type and size
Demographics (gender, race)
Educational Attainment

Output

Average AGI (total AGI / total population)

4 Datasets Used

U.S. Census County Business Patterns (CBP)
U.S. Census American Community Survey (ACS)
- Demographics & Educational Attainment
U.S. IRS SOI Tax Stats - County Data

Results

Model fitted on years 2011 to 2016, evaluated on 2017
91% of predications fall within 30% of the target variable
RMSE of 5.5, R2 value of 0.63
Predictions and coefficients displayed on Tableau for user interaction