Association rule mining using fuzzy logic and whale optimization algorithm

Abstract

Association rule mining (ARM) is a well-known data mining scheme that is used to discover the commonly co-occurred itemsets from the transactional datasets. Two considerable steps of ARM are frequent item recognition and association rule generation. Minimum support and confidence measures are used in the generation of association rules. Many algorithms have been projected by the researchers to generate association rules. Fuzzy logic is incorporated to uncover the recurrent itemsets and interesting fuzzy association rules. In general, huge volume of datasets could be analyzed which in turn needs more number of database scans. In addition to this, all the transactions and items are not required for data analysis. Hence, the first step of this research work uses a dimensionality reduction technique which drastically reduces the size of the data set. This dimensionality reduction technique uses low variance and hash table methods. The proposed algorithm effectively identifies the significant transactions and items from the database. The issues of dimensionality reduction appear when the items in the databases are higher dimension than endure. The proposed algorithm reduces the irrelevant items and transactions from the transactional database. The proposed dimensionality reduction technique dimensionality reduction in transactions and items is compared with the extend frequent pattern (EFP) and intersection set theory EFP and dimensionality reduction using frequency count. Item reduction, transaction reduction, execution time and memory space are the performance factors. Second step proposes fuzzy and whale optimization for frequent item identification and association rule generation. The efficiency of the proposed algorithm is compared with particle swarm optimization genetic algorithm and fuzzy frequent itemset-Miner. Performance metrics used in this step are number of frequent items, association rules generated, execution time and memory required. Experimental results proved that the proposed techniques have produced the good results.