Musbah Hmeda, Aly Hamed H, Little Timothy A
Department of Electrical and Computer Engineering, Dalhousie University, Halifax, Canada.
Heliyon. 2023 Feb 21;9(3):e13903. doi: 10.1016/j.heliyon.2023.e13903. eCollection 2023 Mar.
The stationarity of a time series is an important assumption in the Box-Jenkins methodology. Removing the non-stationary feature from the time series can be done using a differencing technique or a logarithmic transformation approach, but it is not guaranteed from the first step. This paper proposes a new adaptive DC technique, a novel technique for removing a non-stationary time series from the first step. The technique involves transferring non-stationary data into another domain that deals with it as a stationary time series, as it is much easier to be forecasted in that domain. The adaptive DC technique has been applied to different time series, including gasoline and diesel fuel prices, temperature, demand side, inflation rate and number of internet users time series. The performance of the proposed technique is evaluated using different statistical tests, including Augmented Dickey-Fuller (ADF), Kwiatkowski-Phillips-Schmidt-Shin (KPSS), and Phillips Perron (PP). Additionally, the technique is validated by comparing it with a differencing technique, and the results show that the proposed technique slightly outperforms the differencing method. The importance of the proposed technique is its capability to get the stationarity data from the first step, whereas the differencing technique sometimes needs more than one step.
时间序列的平稳性是Box-Jenkins方法中的一个重要假设。可以使用差分技术或对数变换方法来消除时间序列中的非平稳特征，但第一步并不能保证成功。本文提出了一种新的自适应直流技术，这是一种从第一步就去除非平稳时间序列的新技术。该技术涉及将非平稳数据转移到另一个域，在该域中将其作为平稳时间序列来处理，因为在该域中进行预测要容易得多。自适应直流技术已应用于不同的时间序列，包括汽油和柴油价格、温度、需求侧、通货膨胀率和互联网用户数量时间序列。使用不同的统计检验，包括增强迪基-富勒检验（ADF）、 Kwiatkowski-Phillips-Schmidt-Shin检验（KPSS）和菲利普斯-佩伦检验（PP），对所提出技术的性能进行了评估。此外，通过将该技术与差分技术进行比较来验证其有效性，结果表明所提出的技术略优于差分方法。所提出技术的重要性在于它能够从第一步就获得平稳数据，而差分技术有时需要不止一步。
