Figure 1 shows the number of database passes for
the three different types of algorithms: the level-wise
algorithm, Partition, and the sampling Algorithm 1.
Each of the data points in t,he results shown for Algorithm
1 is the average value over 100 trials. Explaining
the results is easy. The level-wise algorithm makes
1<(+1) passes over the database, where K is the size
of the largest frequent set. The Partition algorithm
makes two passes over the database when there are
any frequent sets. For Algorithm 1, the fraction of trials
with misses is expected to be larger than 6 = 0.001,
depending on how many frequent sets have a frequency
relatively close to the threshold and are thus likely
misses in a sample. The algorithm has succeeded in
finding all frequent sets in one pass in almost all cases.
The number of database passes made by Partition algorithm
is practically twice that of Algorithm 1, and
the number of passes of the level-wise algorithm is up
to six times that of Algorithm 1.
การแปล กรุณารอสักครู่..
