Predictive models that establish a linkage between process parameters and part properties have been identified as a high priority research need in Additive Manufacturing. We work with a Multivariate Adaptive Regression Splines (MARS) statistical model to predict the porosity of parts produced using Selective Laser Melting (SLM) process as a function of process parameters. The proposed predictive model is validated through a case study on 17-4 PH stainless steel test coupons manufactured on a ProX 100 SLM system.