A New Algorithm for Mining Frequent Itemsets from Evidential Databases

Mohamed Anis Bach Tobji, Boutheina Ben Yaghlane, Khaled Mellouli.

Association rule mining (ARM) problem has been extensively tackled in the context of perfect data. However, real applications showed that data are often imperfect (incomplete and/or uncertain) which leads to the need of ARM algorithms that process imperfect databases. In this paper we propose a new algorithm for mining frequent itemsets from evidential databases. We introduce a new structure called RidLists that is the vertical representation of the evidential database. Our structure is adapted to itemsets belief computation which makes the mining algorithm more efficient. Experimental results showed that our proposed algorithm is efficient in comparison with the only evidential ARM algorithm in the literature [10].

PDF full paper