Prediction 3-D Velocity for Ecuador by Artificial Neural Network RBF

At present time, Ecuador has a velocity field, but it does not have an interpolation method to calculate the velocities in other points. This study presents a strategy to interpolate the velocities trough the Artificial Neural Network ‑ ANN with a Radial Basis Functions (RBF) type. To exercise this purpose we have used an available dataset in which the geocentric Cartesian coordinates (X,Y,Z) and their velocities (Vx,Vy,Vz) were known. These data were divided in three groups: The first group has been used to training phase; the second group to determinate the capacity of learning of the RBF; and lastly the third group to evaluate of generalization of the RBF by predicting the velocities in these points. In the same manner, we proceeded with the interpolation by using model velocities VEMOS09. Finally, in the test points the velocities differences were calculated, both RBF network as well as those of the VEMOS09. The results obtained demonstrate that interpolation can be better obtained by using a RBF network rather than VEMOS09.