Long-lived ( > 20 days ) sunspot groups extracted from the Greenwich Photoheliographic Results ( GPR ) are examined for evidence of decadal change .
The problem of identifying sunspot groups which are observed on consecutive solar rotations ( recurrent sunspot groups ) is tackled by first constructing manually an example dataset of recurrent sunspot groups and then using machine learning to generalise this subset to the whole GPR .
The resulting dataset of recurrent sunspot groups is verified against previous work by A. Maunder and other Royal Greenwich Observatory ( RGO ) compilers .
Recurrent groups are found to exhibit a slightly larger value for the Gnevyshev-Waldmeier Relationship than the value found by Petrovay and van Driel-Gesztelyi ( Solar Phys . 51 , 25 , 1997 ) , who used recurrence data from the Debrecen Photoheliographic Results .
Evidence for sunspot group lifetime change over the previous century is observed within recurrent groups .
A lifetime increase of 1.4 between 1915 and 1940 is found , which closely agrees with results from Blanter et al . ( Solar Phys . 237 , 329 , 2006 ) .
Furthermore , this increase is found to exist over a longer period ( 1915 to 1950 ) than previously thought and provisional evidence is found for a decline between 1950 and 1965 .
Possible applications of machine-learning procedures to the analysis of historical sunspot observations , the determination of the magnetic topology of the solar corona and the incidence of severe space-weather events are outlined briefly .