Parallel Computer Architecture

2.2 The Parallelization Process 81

ent customers and their transactions and to generate rules for the inference of

customer behavior. For example, the database may store for every transaction the

list of items purchased in that transaction. The goal of the mining may be to deter-

mine associations between sets of commonly purchased items that tend to be pur-

chased together—for example, the conditional probability P(Si/S

2

) that a certain set

of items Si is found in a transaction given that a different set of items S

2

is found in

that transaction, where S

x

and S

2

are sets of items that occur often in customer

transactions. If this probability