The challenge involved predicting revenue between 31-180 days after the first transaction, for about 76.5K new customers based on first transaction details, transaction details for first 30 days and demographic data. The team consisted of data scientists from our Innovation Labs and AI Lab.
The team conducted an extensive feature engineering exercise and evaluated modelling techniques such as LR, RF, ANN, XGBoost. The final submission used a Random Forest model with 24 variables including scores from 2 subproblems that we solved as classification problems.
Read more about the challenge here: https://thedma.org/membership/member-groups-communities/analytics-community/dma-analytic-challenge/