Glosarium AI
Kamus lengkap Kecerdasan Buatan
Temporal Decomposition
Statistical methodology aimed at separating a time series into its fundamental components: long-term trend, cyclical seasonality, and random residuals.
Trend
Component of the time series representing the long-term evolution of the data, smoothed and devoid of seasonal or irregular fluctuations.
Seasonality
Cyclical and predictable component of the time series that repeats at fixed intervals (annual, quarterly, monthly), independent of the trend.
Residual
Part of the time series remaining after extracting the trend and seasonality, capturing random fluctuations and noise.
X-11-ARIMA
Seasonal decomposition procedure developed by the US Census Bureau, combining moving average filters with ARIMA modeling for end-of-series forecasts.
Robustness
Ability of a decomposition method, such as STL, to not be influenced by outliers when estimating components.
Seasonal Period
Number of observations constituting a complete seasonal cycle, essential parameter for decomposition algorithms like STL.
Smoothing Window
Parameter of decomposition algorithms (e.g., STL) controlling the number of observations used for smoothing the trend and seasonality.
Robust STL Decomposition
Variant of the STL algorithm that uses weights to reduce the influence of outliers on the estimation of trend and seasonality.
Seasonal Calendar (Trading Day Effects)
Adjustment made in methods like X-11 to correct the impact of variable day composition (e.g., weekends, holidays) on monthly data.
Henderson Filter
Symmetric moving average filter used in X-11/SEATS procedures to smooth the trend component while preserving inflection points.
Spectral Analysis Decomposition
Approach that decomposes the time series in the frequency domain to identify and isolate cyclical components based on their period.
MSTL (Multiple STL)
Extension of STL allowing the decomposition of series with multiple seasonalities (e.g., daily and annual) by applying the algorithm iteratively.