Bootstrapp resampling methods

Can anyone help me to find some methods (code and teoretical background) of "strategies to cope with the problem of small data sets in the identification of a linear/nonlinear process." Some of the methods in the literature are Marxspline, kubicSpline, Maximum likelihood, Bayesian method etc. method which generate extra data in the boundaries of process dynamics. I need to developed black box model from which the inputs are measured continuoulsy (every minute) and the model output is measured twice per day . I need to generate extra output data with some dynamic (if it exist) resampling method because the process dynamic changes oscilate during 24 hours. My job is to develop some black box model which can predict output during 24 hours.