A new time series robust forecasting approach with application in finance

Subject Areas : Financial Knowledge of Securities Analysis

حمید شهریاری ¹ (ندارد)
نیما شریعتی ² (ندارد)
امیر مسلمی ³ (مسئول مکاتبات)

Received: 2012-01-16 Accepted : 2012-04-03 Published : 2012-12-15

Keywords: Times Series, Autoregressive Model, Outliers, Robust Estimation, Financial Data,

Abstract :

To obtain reliable model for auto correlated and time series data, robust approachshould be considered because outliers and contaminations can have bad effect onparameter estimation of these models. Since most finance data are auto correlated andthey are affected by the previous data, they can be modeled by time series regressionmodels. In this paper, the autoregressive (AR) model is investigated and novel robustprocedure based on filtered S-estimator is proposed to estimate the parameters of ARmodel. This model is used to obtain robust forecasting procedure. We present 148 datagathered from a firm which are related to profit as a numerical example and show theefficiency of the proposed estimation approach. The robust model can forecast moreaccurate than classical model in presence of outlier.