Rare and frequent weighted itemset optimization using homologous transactions: A rule mining approach

The method proposed in this paper deals with weighted itemsets and it considers both frequent and rare itemset mining using Homologous Transactions through the Frequent Pattern Growth Paradigm. Apart from the traditional methods for mining, this method mines out items considering the local interestingness in each of the transactions. Along with the minimum and maximum function, this paper also introduces the average weighting function too to generate the Homologous Transactions. Since average function is used, it is possible to maintain a mid level range of weights, not very high or low. The use of FP-Growth increases the execution time, when compared to Apriori. Two algorithms called, FWIHT (Frequent Weighted Itemset mining using Homologous Transaction) and RWIHT (Rare Weighted Itemset mining using Homologous Transaction), which extract frequent and rare items using homologous transaction are proposed. As an extension, a rule mining task is also done by constructing a matrix instead of FP-Tree and using a measure called cogency. Experimental results prove that the method is more efficient than other mining methods.