AI-ordlista
Den kompletta ordlistan över AI
Lag Features
Features created by shifting time values by k periods to capture past temporal dependencies and autocorrelation patterns in time series.
Rolling Statistics
Statistics calculated on moving time windows such as mean, standard deviation, min and max to capture local trends and data volatility.
Time Series Decomposition
Technique separating a time series into trend, seasonal, and residual components to better understand and model underlying patterns.
Temporal Windowing
Process of segmenting temporal data into fixed or variable size windows for analysis and local feature extraction.
Fourier Transform Features
Features extracted via Fourier transform to identify dominant frequencies and periodic patterns in time series.
Wavelet Features
Features obtained through wavelet decomposition enabling time-frequency analysis to capture local variations and transients.
Autocorrelation Features
Measures quantifying the correlation of a time series with itself at different time lags to identify dependencies.
Seasonal Patterns
Features capturing recurring variations at fixed periods such as daily, weekly, or annual patterns in temporal data.
Trend Analysis
Extraction of features representing the direction and intensity of long-term trends in time series.
Time Delta Features
Features calculated as differences between timestamps or values to measure time intervals and temporal changes.
Cyclical Features
Variables encoding temporal cycles using sine and cosine transformations to preserve the cyclical nature of time.
Time-based Encoding
Techniques transforming temporal information into numerical variables such as day of week, month, quarter, or cyclical encodings.
Moving Averages
Features smoothing short-term fluctuations by calculating averages over sliding windows to reveal underlying trends.
Exponential Smoothing
Method giving more weight to recent observations to create features capturing trends with exponential decay.
Temporal Cross-correlation
Correlation measures between different time series with lags to identify lead-lag relationships between variables.
Time-based Aggregations
Statistics grouped by time periods such as day, week, or month to create summarized features at different granularity levels.
Holiday Effects
Binary or dummy variables indicating holidays and exceptional periods to capture their impact on temporal patterns.
Time Series Segmentation
Division of time series into homogeneous segments based on changes in behavior or distribution for analysis and modeling.