The input file should be a one (value) or two column (timestamp value) ascii file The model is fit from offset "Fit from" to "Fit from + Fit number -1 ". Predictions are made from "Fit from + Fit Number" to "Fit From + Fit Number + Predict Number - 1". At each of these samples, "Number Ahead" predictions are made. The model used is as follows:
A Model is in the form [optional modifier] [required underlying model] Optional Modifiers That Affect Predictors Produced From the Underlying Model ---------------------------------------------------------------------------- REFIT r predictor will refit itself every r data elements AWAIT a predictor will wait for a data elements before fitting MANAGED a r m e v predictor will wait for a data elements before fitting predictor will refit after r data elements predictor will refit if, after m samples, the relative error of one-step ahead predictions exceeds e (avg(abs(obs-pred)/abs(pred)) > e) predictor will refit if, after m samples, the actual error variance of one-step ahead predictions exceeds their predicted variance by a factor of v (variance(error)/predictedvariance > v) Underlying Models ----------------- NONE No model MEAN Long-term mean LAST Last value seen BM p | BESTMEAN p Windowed average, window length chosen to minimize msqerr BMED p | BESTMEDIAN p Windowed median, window length chosen to minimize msqerr AR p Autoregressive model of order p MA q Moving average model of order q ARMA p q Autoregressive moving average model of order p+q ARIMA p d q Autoregressive integrated moving average model of order p+q with d-order difference ARFIMA p d q Fractionally integrated ARIMA model of order p+q. d is ignored and and determined by the model fitting process
http://www.cs.northwestern.edu/~RPS
rps-help@cs.northwestern.edu