TD_GENSERIES4SINUSOIDS generates a result series containing a subset of the sinusoidal elements periodicities (sinusoids).
Sinusoidal elements or periodicities are the components of a time series that exhibit periodic patterns or oscillations over time. A sinusoid is a mathematical function that describes a wave-like pattern that repeats itself over time. The sinusoidal function can be represented as:
y = A * sin(2πft + ϕ)
where:
A is the amplitude of the wave
f is the frequency of the wave
t is the time variable
ϕ is the phase of the wave representing the position of the wave relative to some reference point
In the context of time series analysis, periodicity refers to the pattern of fluctuations in the data that repeats itself over a fixed period. For example, if a data pattern repeats every week, then the periodicity of the data is one week. Similarly, if the pattern repeats every year, the periodicity is one year.
Identifying the periodicities in a time series is used in forecasting, anomaly detection, and signal processing. By identifying the periodicities, you can model the underlying patterns in the data, and use them to make predictions or detect anomalies.
- Use TD_LINESPEC or TD_POWERSPEC with FREQ_STYLE parameter set to K_PERIODICITY to determine the periodicities in the series.
- Use a SELECT statement on the output similar to the following to view the periodicities.
SELECT TOP 5 * FROM LineSpecResults ORDER BY SPECTRAL_DENSITY_field DESC;
- Use TD_GENSERIES4SINUSOIDS with the PERIODICITIES parameter and a comma-separated list of periodicities to exclude from the data set.
- Use TD_BINARYSERIESOP to subtract the generated series from the original series using MATHOP(SUB) parameter.
- Use TD_POWERSPEC to verify that target periodicities have been removed from the original series.