SSACD - Singular Spectral Average Convergence/Divergence
This is an analogue of the MACD indicator based on singular spectrum analysis (SSA).
SSA is an effective method to handle non-stationary time series with unknown internal structure. It is used for determining the main components (trend, seasonal and wave fluctuations), smoothing and noise reduction. The method allows finding previously unknown series periods and make forecasts on the basis of the detected periodic patterns.
The recognized smoothed price fluctuations are not lagging in time relative to the source series as opposed to MACD, which uses moving averages. Accordingly, this indicator has solved the MACD problem of lagging to more accurately and synchronously reflect the variability of the price series behavior. The resulting forecast for the identified "fast" and "slow" price fluctuations takes into account the aggregate analysis of various detected factors, which form the "wave" behavior of the data series, and it can be used to reduce risks in a strategy.
The characteristic behavior, as well as the signals and interpretation of the indicator correspond to the same properties of the linear MACD.
- Algorithm — the forecast algorithm (1 - vector, 2 - recurrent)
- N: Data fragment — the length of the analyzed price series.
- Time-dependent lag — determines the window of data "influence" (N/2, ..., N/4).
- FastTrend High frequency limit — a filter parameter for identifying a fast oscillating equivalent of the MA.
- SlowTrend High frequency limit — a filter parameter for identifying a slow oscillating equivalent of the MA.
- Signal SMA period — the smoothing period of the difference between the fast and slow moving averages
- Data preparation — the method of preparing data for the analysis
- Forecast preparation — the method of preparing data for the forecast
- Recalculate period — indicator refresh interval (c)
- Predictable Points — the number of prediction points.
- BackwardShift — a shift of the analyzed fragment deeper into the history. Configuration of the model and forecast based on known data.
Explanation to the set of parameters
The influence window limits the number of past points which affect the current price. A reasonable value if s = 1 or 2.
High frequence limit for defining MAs indicates the weight of the high-frequency noise in the total variance of the price series (in percentage). For the fast oscillating MA it should be equal to 0.5 - 1.5, for the slow one 1.5 - 4. All oscillations, whose weight does not exceed this level will be filtered out.
BacwardShift shifts calculations along the price series in order to compare the forecast with the known prices and to select the indicator parameters.
Data preparation - modification of the price series for another sequence in order to improve the forecast result on a certain timeframe.
P.S. Trend visualization along the price chart is available using the SSA Trend Predictor indicator.