Comparative analysis of properties of weakening buffer operators in time series prediction models

Reducing the negative influence of stochastic disturbances in sample data has always been a difficult problem in time series analysis. In this paper, three new fractional weakening buffer operators are proposed, and then some desirable properties of these proposed sequence operators are investigated. Their potential effect in smoothing unexpected disturbances while maintaining the normal trend in sample series is analyzed and compared with other widely used sequence operators in time series modeling. Results of theoretical and empirical research show that the proposed novel fractional weakening buffer operators are effective in improving the development pattern analysis of time series in disturbance scenarios, while also avoid too subjectively weighting experimental data from collected samples. The robust of the proposed operator-based prediction algorithm against noise effect is tested in five different types of noise scenarios. Result of empirical study demonstrates that the proposed method improves the series prediction performance and it also improves the robustness of corresponding forecasting algorithms. These unique properties of the proposed weakening buffer operators make them more attractive in time series analysis.