Here we show a few examples of applying FindGraph
for analysis and forecasting of various time series from a published graph.
First we capture the area of interest and digitize it.
Open finance.yahoo.com page in Explorer, start FindGraph and select menu item "Add data from Screen".
Move the framework on screen to select the rectangle of digitizing and to capture the area of interest.
The Wizard of Digitization will help you to digitize the data. Graphs are handled automatically with fixed step to extract time series.
We selected step value 0.1 mm to get about 500 data points.
Next we apply SSA to reconstruct and to forecast time series.
A SSA (Singular Spectrum Analysis) method can be used for time series prediction.
In dialog select window length value 10. To add components select 4 first items checked.
Select 70 points to forecast.
You can visually compare prediction results with input time series,
and to vary parameters to find optimal for prediction.
An examples of applying FindGraph for analysis of EUR vs. USD time series.
An examples of applying FindGraph for analysis and forecasting of currencies exchange rates
based on a graph published.
To compare short term and long term forecasts, we prepare time series testing data based on formula:
y = 0.2x+sin(x)+random. Next we apply SSA, ARIMA, GMDH, and cyclic extraction method to forecast time series.
Results are compared with y = 0.2x+sin(x).