Trust: A new objective measure for symmetric association rule mining in account of dissociation and null transaction
Subrata Datta, Kalyani K Mali · 2017
Association rule mining from the large transactional database is one of the interesting and challenging paradigms in knowledge discovery. An objective measure is a key tool for the measurement of interestingness in between two patterns. Association rules extracted using traditional objective measures may have high dissociation which is by nature opposite of association. Dissociation in between two patterns refers to the presence of one pattern and absence of another in the same transaction. High dissociation indicate less association and vice versa. So an association rule with high dissociation makes no sense. This is caused due to the negligence of dissociation between patterns while formulating the objective measure. In this paper we have introduced a new objective measure called trust which is formulated in the context of dissociation. Moreover trust is a symmetric objective measure and takes account of impact of null transactions over positive association rules. Experimental studies with synthetic and real datasets demonstrate the validity and effectiveness of the proposed objective measure.