Abstract—A lot of algorithms performing Frequent Itemsets Mining (FIM), however, some of the glitches in the algorithms still require attention, particularly when the mining process involves a high dimensional dataset. The Directed Acyclic Graph in High Dimensional Dataset Mining (DAGHDDM) is a graph-based mining algorithm that represents itemsets in the complete graph before FIM takes place. Nevertheless, the construction of complete graph creates unnecessary edges and makes the search space large and affects the overall algorithm performance. This research aims to speed up the searching process by creating relevant edges in the graph to reduce the search space by rearranging the items using the common prefix rowset. We proposed a novel frequent itemsets mining using row enumeration approach on graph based structure called Frequent Row Graph Closed (FRG-Closed). Designing the FRG-Closed involves new data structure creation known as Frequent Row Graph (FR-Graph). We performed the experiments to compare the performance of FRG-Closed with DAGHDDM algorithm. The result of the experiments revealed the FRG-Closed capability to mine the frequent closed itemsets faster than its counterpart, DAGHDDM algorithm. Moreover, the FRG-Closed is also able to handle lower minimum support compared to the DAGHDDM for a larger transaction.
Index Terms—Data mining, graph theory, high dimensional, frequent itemset.
Shuzlina Abdul-Rahman is with the Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, 40450 Shah Alam, Selangor, Malaysia (email: email@example.com).
Cite: Mohammad-Arsyad Mohd-Yakop, Shuzlina Abdul-Rahman, and Sofianita Mutalib, "Novel Row Enumeration Approach of Graph-Based Frequent Itemsets Mining," International Journal of Machine Learning and Computing vol. 8, no. 4, pp. 324-330, 2018.