Self-Organizing Feature Map Based Data Mining

Shangming Yang, Zhang Yi

Abstract: In data mining, Apriori algorithm for association rules mining is a traditional approach. However, it takes too much time in scanning database for finding the frequent itemsets. In this paper, based on SOM clustering, a novel algorithm is introduced. In this algorithm, each transaction is converted to an input vector, SOM is employed to train these input vectors, from which we achieve the visualization of the relationship between the items in a database. The time efficiency and the visualized map units make the proposed approach a particularly attractive alternative to current data mining algorithms.