Toddler stunting is a chronic nutritional problem caused by one of which is inadequate nutritional intake in infants. Bojonegoro is a district in East Java province where the incidence of stunting under five is still common. In this study, we will compare the incidence of stunting under five in Bojonegoro district using GWR and MARS. GWR is able to model cases for each region spatially, while MARS is able to model cases without considering the pattern of relationships between predictor variables and response variables. The best GWR model selected the smallest CV and MSE values ​​and the largest R-Square values. The best model of the Adaptive Bi-Square kernel function was obtained with a bandwidth of 28, the value of CV=2.4635, MSE=0.8620, and R-Square=0.8734. The best MARS model selected the smallest GCV value and the largest R-Square. From the combination of MO, BF and MI values, the best model was obtained at BF=24, MI=1 and MO=1 with GCV=1,29144 and R-Square=0,841 values. Comparing the two models, the MARS model is better because the R-Square value of the GWR model is greater than the MARS although the numerical value is not much different, while the MSE value of the MARS model is smaller than the MSE value of the GWR model with a much different difference. The results of this study will provide knowledge in the form of a regression model for the Bojonegoro district health office in predicting predictor variables that affect the incidence of stunting under five in Bojonegoro district.