A Stochastic Algorithm of Frequent Set Search for Mining Association Rules

Data mining is discovery of unknown, nontrivial, practically useful and easy to interpret knowledge in chaotic data. The information, found by application of data mining techniques, is unknown in advance. Knowledge is described by relationships of new features that distinguish one attribute value from other set attributes. The new knowledge set must be applied to new information with some degree of reliability. Current algorithms for finding association rules require several passes over the analysed database. The paper presents a stochastic algorithm for mining association rules in large data sets. Our stochastic algorithm reduces the database activity considerably, because the database is analysed only once. The algorithm allows us to measure two significant criteria, i.e. time and accuracy. We analyse a large database of transactions. Each transaction consists of items purchased by a customer in a visit. The algorithm yields conclusions about association rules using the analysis of randomly selected subsequences. Our experiments show that the proposed algorithm can find association rules efficiently in only one database scan.