PMML in Simulation

摘要

Predictive models use a set of inputs to predict an outcome. In cases where the input values are not known with certainty, Monte Carlo simulation can be used to generate input values based on specified probability distributions for the uncertain inputs. The simulated values are then used in the predictive model to generate an outcome. The process is repeated many times, resulting in an approximation to the distribution(s) of the target (or targets) of the model that can be used to answer questions about the likelihood of various outcomes. Predictive Model Markup Language (PMML) is an industry standard for representing statistical and data mining models in XML. It is a very convenient way to represent model information for use in a Monte Carlo simulation. This paper describes how PMML enables a seamless transition from traditional modeling to modeling in the presence of uncertainty through Monte Carlo simulation.