This is the material from Gazprom Neft SmartOil Contest which took place on boosters.pro platform (https://boosters.pro/gazprom_smartoil_contest). The object of this contest was to predict the amount of oil production per oil well for the next 6 months (in tons). The evaluation metric is MAE.
The repo consists of an IPython notebook with some data preprocessing techniques and a general XGBoost model (w/o linear regression model from the final submission), a presentation (RUS) of the author's view on this problem and the overall approach.