Abstract: The Least Absolute Relative Error Estimation Method (LAREEM) was compared with the most important method of the Least Squares (LSM). The autoregression models, also with periodic components, for the unemployment rate in Poland in years 1990-2010 are tested. The algorithm which is used to calculate the structure parameters is based on subgradient method. The purpose of these investigations was to find the best model to predict as well as possible the unemployment rate in the future. Some models with periodic components can predict the unemployment rate in Poland for next month with relative error less than 3%.

Keywords: unemployment rate, linear autoregression, relative error

Tagi: linear autoregression | relative error | unemployment rate