CAJ | 학술논문

针对混沌时间序列自适应预测方法在多步预测中预测器系数无法调节的问题,根据混沌时间序列的短期可预测性及自适应算法的自适应跟踪混沌运动轨迹的特点,提出了一种自适应多步预测方法.在多步预测中,该方法根据已知样本得到对将来值的预测值并能自适应调节滤波器系数.仿真结果表明此方法的多步预测性能明显好于自适应预测方法的多步预测性能.将此方法应用于对股票数据的预测,得到了较好的预测结果.
침대혼돈시간서렬자괄응예측방법재다보예측중예측기계수무법조절적문제,근거혼돈시간서렬적단기가예측성급자괄응산법적자괄응근종혼돈운동궤적적특점,제출료일충자괄응다보예측방법.재다보예측중,해방법근거이지양본득도대장래치적예측치병능자괄응조절려파기계수.방진결과표명차방법적다보예측성능명현호우자괄응예측방법적다보예측성능.장차방법응용우대고표수거적예측,득도료교호적예측결과.
Based on the short-term predictability of chaotic time series and the adaptive tracking chaotic trajectory of adaptive algorithm, a novel adaptive multi-step-prediction method is proposed to resolve the problem of adjusting filter＇ s parameters of the adaptive prediction method during multi-step prediction in this paper; and this method is used to predict typical chaotic time series and chaotic stock data. During multi-step prediction this method can adjust filter＇s parameters and get the multi-step prediction value using the data we know. Simulation results show that： this adaptive multi-step-prediction method can be successfully used to make multi-step predictions of typical chaotic time series and stock data, and this method＇s multi-step-prediction performance is better than the adaptive prediction method＇s.