7.1.3 Effect of minSup
2500Figure 9 shows the influence of minSup on the runtime
2000when using different densities. The runtime of Dynamic
Dynamic
directly correlates with the size of the dynamic computation
1500matrix (Figure 5). A low minSup value leads to few matrix
1000Runtime [ms]rows which need to be computed, while a high minSup value
500Dynamic+P
leads to a slim row width (see Figure 5). The total number of
0matrix cells to be computed is minSup ∗(|T | −minSup + 1),
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1with a maximum at minSup = |T|+1 2 . As long as the minSup
Minimum supportvalue is below the expected support value, the approach with
(b) Density = 0.5
(a) Density = 0.2
pruning shows similar characteristics; in this case, almost all
item(sets) are expected to be frequent. However, the speed-
up due to the pruning rapidly increases for minSup above
Figure 9: Runtime evaluation w.r.t. minSup.
this break-even point.
Management of Data (SIGMOD’94), Minneapolis, MN,
Bạn đang xem 7. - BAI 3 URBAN PLANNING