Various aspects include vectorization approaches for model-based mask proximity correction (MPC). In some cases, a computer-implemented method includes: assigning a set of vectors to geometry data describing at least one mask for forming an integrated circuit (IC); adjusting a statistical predictive model of the at least one mask based upon the set of vectors and the geometry data; predicting an adjustment to the at least one mask with the statistical predictive model; and adjusting instructions for forming the at least one mask in response to a predicted mask result of the statistical predictive model deviating from a target mask result for the at least one mask.