Design of an Error Estimation Algorithm for Time Series Data Prediction in Attitude Controlled Systems

Abstract

This research presents an error estimation approach with the combination of traditional multilevel techniques used to minimize errors for an accurate prediction and to investigate the behavior of such an algorithm for a satellite. The traditional techniques mentioned above are a combination derived from multiple regression techniques and perform a case study for data analysis. A linear plot can easily be predicted, however if a system tends to deviate toward non-linearity the overall result derived from such an algorithm can be non-reliable since each value would depict a completely different output at different levels. The stability of the fixed regression derived is used to determine the accuracy of the system.

Keywords

• neural networks
• regression analysis
• mean absolute error
• vector auto regression