论文部分内容阅读
针对飞机平均故障间隔飞行时间(MFHBF)指标值的预测问题,提出一种基于时间序列分解的组合预测方法。首先利用STL(Seasonal and Trend Decomposition using Loess)方法分解出MFHBF指标的长期趋势项和季节变动项,然后以灰色模型预测指标的长期趋势,以BP神经网络和支持向量机回归组合模型分别预测指标的季节变动,根据误差权重计算季节变动的加权值,最后以加法模型合并趋势和季节的预测值获得最终结果。利用服务点积累的指标数据对方法进行检验,与单独使用支持向量机回归预测得到的结果相比,平均绝对误差由45%减小至21%,证明该方法能够有效提高预测精度,为保障人员提供可信的指标预测结果。
Aiming at the prediction of MFHBF index value, a combined forecasting method based on time series decomposition is proposed. Firstly, the long-term trend and seasonal variation of MFHBF are decomposed by using the method of seasonal and normal decomposition (STL). Then the long-term trend of the index is predicted by the gray model. The combined model of BP neural network and support vector machine regression is used to predict the long- Seasonal variation, weighted according to the error weight of the seasonal changes, and finally the results of the merger trend and the season forecast by the additive model to obtain the final result. Compared with the results of SVM regression alone, the average absolute error is reduced from 45% to 21%, which proves that this method can effectively improve the prediction accuracy, Provide credible indicator forecast results.