Adding the constant > 0
to the regret value assures that the selection probability for each activity in the
decision set is greater than zero and thus every schedule of the population can
be generated. With the choice of the parameter the amount of bias can be
controlled. A high will cause no bias and thus deterministic activity selection
while an of zero will cause maximum bias and hence random activity selection.
Kolisch (1995) found out that, in general, = = 1 will provide good
results. Drexl (1991) uses = mini2Dg v(i). Schirmer and Riesenberg (1997)
propose a modied regret based biased random sampling (MRBRS) where
is determined dynamically. Experimental comparisons performed by Kolisch
(1995) as well as Schirmer and Riesenberg (1997) revealed (modied) regret
based biased random sampling as the best sampling approach.