Mining Rare Association Rule Arnab

Abstract — Pattern mining is a data mining method that involves finding existing patterns in data. The main objective is to uncover the hidden regularities in the data. Until now most of the research in pattern mining has been exclusively focused on frequent item set, however in some situation it is interesting to find patterns that are rare instead of frequent or patterns that reflect a negative co relation between items. These patterns are referred to as rare and negative patterns. For example in jewellery sales data, sales of diamond watches are rare however patterns involving the selling of diamond watches could be interesting. Rare items occur rarely but are of special interest as they may highlight exceptional behaviour in the data which is likely of interest. It tries to capture patterns that involve events that are unusual in the data set. This research proposes that in some domain such as medical databases, computer security, rare items are of higher importance than frequent one. This research investigates other kind of knowledge that is not usually discovered using frequent item-set mining i.e. rare item-sets and non-present item-sets. In order to mine rare and non-present item-sets, we methodologically extract the principles behind Apriori approach, and then generalize it in such a way it can not only be applied to mine frequent item-sets but also to mine other categories of patterns such as rare and non-present item-sets. Our goal is to apply the principle of the Apriori approach used for frequent item-set mining and apply it to rare and non-present item-set mining. The proposed approach is able to find the whole rare item set and at the same time pruning out the zero itemset.We will be using the concept of minimal rare item set (MRI) to generate strong Rare Association rules with high confidence and low support. The MRI will act as a generator, based on which valid rare association rule will be generate