An Adjacency matrix-based Multiple Fuzzy Frequent Itemsets mining (AMFFI) technique

Abstract

Recently, discovering helpful information from a database consisting of transactions has been a critical research topic. Several frequent itemsets mining for association rule mining, algorithms that can only handle binary databases have proposed. Transactions using numerical values, on the other hand, are ubiquitous in real-world applications. Thus, with reference to the quantitative transactional database, several algorithms were developed and “fuzzy frequent itemsets” (FFI) were discovered. Most of them just consider the term having maximum cardinality. As a result, the number of fuzzy regions processed is equal to the number of original elements. Multiple fuzzy zones of an item, on the other hand, give a better result for making a correct decision. This study presents an AMFFI-miner (Adjacency matrix-based Multiple Fuzzy Frequent Itemsets) for discovering multiple FFIs out of a quantitative transactional database. An adjacency matrix and fuzzy-list structure were designed to find multiple FFIs by scanning database only once and generates less number of candidate itemsets. Join two nodes if its co-occurrence between two fuzzy linguistics terms satisfies minimum support threshold by finding the co-occurrence between two fuzzy linguistics terms directly from the adjacency matrix, thus reducing the number of nodes joining and speeding up discovering multiple FFI. Experiments carried out to compare the suggested method's performance to that of existing methodologies based on running time, memory utilization, and the number of nodes joining.