A better method was a Boosted Decision Tree (BDT), a type of machine learning algorithm that learned to differentiate between "constant" and "microlensing" light curves.