iFP-growth: Um Algoritmo Incremental para Determinar Regras de Associação

Association rule mining has been widely used in decision support systems and web site personalization. Many of these systems are characterized by frequent data additions, which inevitably determines the mining results’ re-evaluation. In such a scenario, the use of traditional algorithms which mine the whole database each execution may be wasteful or even impracticable. In this paper we present iFP-Growth, an algorithm that from the original set of association rules and additional data, calculates the resulting association rules. This algorithm is an extension of the FP-Growth algorithm and distiguishes from the later not only for its incremental nature, but also for the optimization of computing resources usage. The algorithm was implemented in Java and evaluated for both synthetic and real data, showing its advantages and the maintenance of the original algorithm’s properties.