TD_BINARYSERIESOP performs a point-wise mathematical operation on two time series of equal size. The first input is referred to as the series 1 (primary input). The second input is referred to as series 2 (secondary input). The mathematical operation is addition, subtraction, multiplication, or division.
Common uses of TD_BINARYSERIESOP are:
- Subtracting trends from a time series to create a model from it.
- Restoring trends to a time series before using the model for forecasting.
- Using the function as a building block to formulate more complex functions.
In time series data, a trend is a long-term pattern in the data that changes gradually over time. Time series datatrends for various reasons, including the following:
- Seasonality: Data has periodic trends due to seasonal effects, such as sales of winter clothing during colder months.
- Economic cycles: Data trends as the economy goes through cycles of growth and contraction. For example, the demand for certain goods and services decrease during a recession.
- Population changes: Population changes cause trends in time series data. For example, if the population of a city is increasing, the demand for housing may also increase.
- Natural disasters: Data changes during natural disasters. For example, a hurricane or earthquake can cause a sudden drop in sales or production, followed by a recovery period.
Unwanted trends may represent noise or other factors that interfere with the analysis of a wanted trend. Unwanted trends make it difficult to detect and analyse the underlying pattern in the data, and may need to be removed or accounted for. For example, a stock analyst may need to remove the effect of daily fluctuations in stock prices to analyse the long-term trend of the stock's performance.
Removing unwanted trends from a time series before analysis has the following advantages bu improving:
- Forecasting accuracy
- Pattern identification
- Model stability
The following procedure is an example of how to use TD_BINARYSERIESOP to convolve two series with digital signal processing:
- Use TD_DFFT on series 1 (primary input) and series 2 (secondary input) to two ARTs named dfftRes1 and dfftRes2, respectively.
- Use TD_BINARYSERIESOP to do point-wise multiplication using the two result sets,dfftRes1 and dfftRes2, and place the result in an ART named freqRes.
- Use TD_IDFFT on freqRes to get the convolved result of the two series.