| Summary: | Linear programming (LP) has been applied extensively in the areas of business and operations management. However little has been in exploring the use of linear programming as a forecasting tool.
This dissertation attempts to explore the application of Linear programming in the area of Forecasting. A study is conducted to find out whether simple linear models that are defined with the use of LP can provide adequate results in relation to nonlinear and simple known models. Additionally, the hypothesis that the combination of various forecasting methods improves forecasting accuracy is also tested.
A data sample of 60 time series is been used to test the forecasting accuracy of 12 linear, nonlinear and simple models. The forecasting accuracy of these models is checked through three different conditions: a) Short term b) Intermediate and c) Long term forecast horizons.
The results showed that linear models and LP can be used as a forecasting tool, since they can provide adequate results. Especially in the short and long term horizons, linear models that were defined with the use of LP demonstrated very good results compared to a number of nonlinear and simple models.
Finally this study confirmed that the combination of methods is a good optimization strategy, while a number of other relative findings agreed with the results of the M- Competitions (Makridakis et al. 1982, 1993, 2000).
|
|---|