Modelling of reservoir-induced earthquakes: a multivariate adaptive regression spline

This paper employs a multivariate adaptive regression spline (MARS) for the prediction of the maximum magnitude (M) of reservoir-induced earthquakes based on reservoir parameters. MARS is a non-parametric adaptive regression procedure. It has the capability to determine the important sequence of inputs for the output. The comprehensive parameter (E) and maximum reservoir depth (H) are used as inputs for the MARS. The developed MARS gives an equation for the prediction of M. A comparative study is carried out between the developed MARS and other models and the results show that the developed MARS is a robust model for the prediction of M.