A computer-implemented method of predicting an output of a model includes receiving output data from a first run of a target predictive model. One or more additional sets of output data is received from runs of one or more alternate predictive model. The alternate predictive models predict parameters that relate to the output data of the first predictive model. A future output of a second run of the target predictive model is predicted based on the output data from the first run of the target predictive model and one or more sets of additional output data from runs of the alternate prediction model.