Running time analysis

• We tested SPAM against the SPADE and PrefixSpan algorithm, which also mine frequent sequences.

• Sequence data was generated using the AssocGen data generator, which allowed us to specify several parameters concerning the size of the dataset (e.g. # of customers, avg. # of transactions per customer, avg. # of items per transaction)

• We ran tests that varied the minimum support as well as tests that varied these dataset parameters.

• SPAM runs faster than both SPADE and PrefixSpan on the datasets we tested by up to an order of magnitude

• SPAM performs especially well when the the datasets contain numerous long sequences.

Slide Links:

Slide Comments: