Predicting dangerous seismic events in coal mines under distribution drift

We describe our submission to the AAIA'16 Data Mining Competition, where the objective is to devise a reliable prediction model for detecting periods of increased seismic activity in coal mines. Our solution exploits a selective naive Bayes classifier, with optimal preprocessing, variable selection and model averaging, together with an automatic variable construction method that builds many variables from time series records. One challenging part of the competition is that the input variables are not independent and identically distributed (i.i.d.) between the train and test datasets, since the train data and test data rely on different coal mines and different times periods. We apply a drift-aware methodology to alleviate this problem, that enabled to get a final score of 0.9246 (team marcb), less than 0.015 from the challenge winner.