Built a machine learning model to predict income from worker-specific characteristics using Current Population Survey data. Joint submission with Hasan Tosun to the MEBDI ML competition (2018). The goal of the competition was to improve on the Root-Mean-Square error of the Mincer Regression framework of income. Our algorithm improves the prediction error on test data by 35%.
The Mebdi_Report_Dalani_Tosun.pdf
contains a detailed report of our submission entry. The incomeprediction.ipynb
file provides the Python code to the second part of the competition. Variables_Dalani_Tosun.xlsx
contains information on the variables used.