Combining forecasts using recursive equal weighting and linearprogramming

摘要

Two combining methods, called recursive equal weighting (REW) andlinear programming (LP), respectively, are introduced. Their forecastingperformance is compared with other combining methods. The proposed REWmethod not only reaches a forecasting accuracy comparable to that of thetheoretically optimal combining methods, but also provides some valuableinsights into the methodology of combination, i.e., the combination ofcombined forecasts can further improve the forecasting performance. Theoccurrence of outliers is shown to lead to inaccuracy in the ordinaryleast squares (OLS) solution, while the proposed LP method caneffectively eliminate their effects