Original scientific paper

Measurement of filters’ efficiencies and application of NNSFDI method

Melike Bildirici
Selcuk Alp

Abstract

The study aims to propose the neural network filter based on NNSFDI method as an alternative filter vis-a-vis to frequently applied filters in economics; namely, Baxter-King, Hodrick-Prescott, Christiano and Fitzgerald and Kalman filters. In this paper, it was used two different data which consist of the annual unemployment rates for 1923 - 2008 periods and the monthly inflation rates for 1964:02 - 2009:07 periods. The performance of the new method proposed and the main stream filters were, in particular, evaluated based on the annual and monthly data. The empirical findings suggest that the newly proposed NNFSDI model provided better forecast results compared to Kalman, HP, BK and CF filters for different data sets when evaluated in the light of different error criteria such as MSE, RMSE and MAPE.

Keywords

Baxter-King filter; Hodrick-Prescott filter; Christiano and Fitzgerald filter; Kalman filter; NNSFDI model; adaptive threshold algorithm

Hrčak ID:

97020

URI

https://hrcak.srce.hr/97020

Publication date:

1.12.2012.

Article data in other languages: croatian

Visits: 1.298 *