faqs.org » Abstracts Index » Abstracts » Business, general

An empirical evaluation of alternative forecasting combinations

Article Abstract:

Forecasting models may be combined to improve accuracy even at only a slightly higher cost. Models are combined here selectively rather than using them all in a single set, examining six new model variations and comparing them to three previously studied sets. Time horizons are analyzed and error measurements are taken to the mean square error. Results suggest using the model combination that picks the best three to five models out of ten and ranks the models based on the inverse proportion of their individual accuracy as measured against mean square error.

Methods, Decision-making, Decision making

---

Multipass seasonal adjustment filter

Article Abstract:

Researchers have devised a new seasonal decomposition algorithm and a state-space seasonal time series model based on the Kalman filter. A multipass filtering method provides the basis for the new seasonal adjustment and decomposition algorithm. Real time series were used to empirically compare the multipass seasonal adjustment filter to the Census X-11 procedure. Results indicate that the proposed method is superior for all time series in the study

Business forecasting, Decomposition method

---

A state space modeling approach for time series forecasting

Article Abstract:

Forecasts of autocorrelated time series and related recursive estimation processes are developed, using a stochastic filtering method. The models developed are state space models, applicable to both seasonal and non-seasonal time series, and employ Kalman filters and techniques borrowed from Brown's adaptive smoothing theories. On-line time series forecasting, problem estimations and proofs, and initial value determinations are also discussed.

Case studies, Stochastic analysis, Kalman filtering

Some parts © 2012 Advameg, Inc.