Time series analysis and forecast, Noise reduction

Examples: Time series analysis

Noise reduction

The 'Caterpillar'-SSA is a technique applicable to separation of the signal from the noise (noise reduction). Natural decomposition of the time series is constructed on the base of the unique parameter (the window length). Grouping regular decomposition components, you can extract the signal from the time series.

Figure 1.5: White dwarf: smoothed series

Figure 1.6: White dwarf: residuals

Data: "White dwarf"
Description: 618 point measurements of the time variation of the intensity of the white dwarf star PG1159-035 during March 1989
Source: The Santa Fe Time Series Competition Data
Parameters: window length = 100
Figure 1.5: the time series component reconstructed from the leading 11 eigentriples (red line) and the initial time series
Figure 1.6: residuals of the reconstruction from the leading 11 eigentriples

Previous example | Next example