Introducing Flow Field Forecasting

A review of flow field forecasting: A high‐dimensional forecasting procedure

“…According to the methodology presented in (Frey and Caudle (2011); Frey and Caudle (2013)), every time series has a set of histories that are based on overlapping segments of the time series. In order to differentiate, s…”

“…A central aspect of FF forecasting is that this methodology forecasts the changes in observations as opposed to the levels themselves. Although this is equivalent to predicting future levels of the time series, it emphasizes that the underlying generative process is a set of stochastic differential equations (Frey & Caudle, 2011, 2013).…”

“…First introduced in 2011 (Frey & Caudle, 2011), FF forecasting was developed initially for the U.S. Department of Energy's (DOE) dedicated internet ESnet (ESnet, Energy Sciences Network, n.d.). In this review, we will provide the history of FF forecasting and explain the FF forecasting methodology.…”

“…In this review, we will provide the history of FF forecasting and explain the FF forecasting methodology. For more detailed and comparison studies, we refer the reader to the published work on this subject (Caudle, Fleming, Pyeatt, & Hoover, 2019; Caudle & Fleming, 2016; M. Frey & Caudle, 2011; Frey & Caudle, 2013).…”

Flow field forecasting for univariate time series

“…Subsequent steps II and III of flow field forecasting are based solely on this spline smoother. The collection of information in the spline smoother is called the process data skeleton [14], and because the size of the skeleton is only loosely related to and much smaller than the size of the original data record, this stage scales large data sets to a more manageable size and supports computation in cases of very limited resources. Standard penalized spline regression involves a numerical search for the appropriate amount of smoothing.…”