Combined Forecasting - Linear Programming
Trinh, Hoang Quan (2012) Combined Forecasting - Linear Programming. [Dissertation (University of Nottingham only)] (Unpublished)
A large literature has accumulated over the years regarding the application of Linear programming in the area of forecasting. This study investigates whether LP models is an attractive technique for forecasting in terms of accuracy improvement. In addition, the performance of the individual forecasts is also tested and compared with traditional combinations methods in order to examine whether the combined forecasts improve the performance of the individual forecasting techniques. The primary conclusion of this research is that LP models can be used as a forecasting tool. They outperform the traditional combination techniques as well as individual techniques in general. In addition, from the findings, it can be concluded that forecast accuracy can be substantially improved through the combination of multiple individual forecasts. The limitation of the study is the small number used of time series, forecasting models as well as forecasting error measures in running experiments. Recommendations for future research directions include (1) an increase of above mentioned components in experiments, (2) study of application of linear programming and combined forecasts in the areas of business and operations management, (3) development of combining forecasts in an optimal way, and (4) automation of time series classification and analysis process.
Actions (Archive Staff Only)