| Key Words |
Time series forecasting; differential evolution; information reduction |
| Abstract |
Time series forecasting is one of the high researched fields. Accurate forecasts influence many daily activities of the people and the businesses. Many difficult decisions are taken augmented by particular mathematical models introduced by time series forecasting researches. With data organized as time series many preprocessing calculations can be done before this data to be supplied at the input of the forecasting model. A common preprocessing manipulation is the removal of the trend from the time series. This type of calculation is done by linear regression and mathematical subtraction of the linear component from the original data. Usually in the time series there is seasonality. By calculation of the coefficients for sinusoidal harmonics seasonality also can be subtracted from the original data. In this research a differential evolution optimization is proposed in order the trend and the sinusoidal harmonics to be removed. By such transformation the forecasting complexity of the time series is decreased. |
| Article PDF | Download article (PDF) |
