Abstract
Pattern discovery techniques, such as association rule discovery is one of the fundamental problem in data mining. Usually the task is limited to positive rules of the form X → Y when X and Y are susbets of items. To enlarge the knowledge discovery from data. Many works pointed out that other rules can be mined linking the present items in transactions with the missing ones designed as negative rules. To mine the most relevant negative rules, the mining task of negative association rules is often coupled with new measure such as lift or conviction to limit the set of extracted association rules.In this work we address the problem of mining strong negative rules by extending the SAT-Based approach proposed in [1]. We show that the conviction constraint leads to a non-linear constraints that have to be managed efficiently to prune the search space. Experiments results explore the efficiency of our new approach.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have