developed, e.g., by taking into account the number of
candidates on the level .and whether the number of frequent sets per level is growing or shrinking. The observations
made from Figure 2 hint that the lowered
threshold can be set in the start-up to roughly correspond
to the desired probability of a miss, i.e., for
∆= 0.01 the lowered threshold could be set as for
δ = 0.01