A proposed novel adaptive DC technique for non-stationary data removal

Area of Science:

• Time Series Analysis
• Econometrics
• Statistical Modeling

Background:

• Stationarity is a critical assumption for Box-Jenkins time series forecasting.
• Existing methods like differencing or logarithmic transformations may require multiple steps to achieve stationarity.
• Non-stationary time series present challenges for accurate forecasting.

Purpose of the Study:

• To introduce a novel adaptive DC technique for efficiently removing non-stationarity from time series data.
• To demonstrate the technique's ability to achieve stationarity in a single step.
• To compare the proposed method's performance against the traditional differencing technique.

Main Methods:

• A new adaptive DC technique is proposed, involving data transformation into a different domain.
• The technique was applied to diverse time series: fuel prices, temperature, demand, inflation, and internet users.
• Performance was evaluated using Augmented Dickey-Fuller (ADF), Kwiatkowski-Phillips-Schmidt-Shin (KPSS), and Phillips Perron (PP) tests.

Main Results:

• The adaptive DC technique successfully transformed non-stationary data into stationary data in the first step.
• Statistical tests (ADF, KPSS, PP) confirmed the stationarity of the transformed data.
• The proposed technique showed slightly superior performance compared to the standard differencing method.

Conclusions:

• The adaptive DC technique offers a more efficient approach to achieving time series stationarity compared to differencing.
• Its ability to attain stationarity in one step simplifies and potentially improves forecasting accuracy.
• This method holds promise for various applications involving economic and environmental time series data.