An optimized FP-growth algorithm for discovery of association rules

Association rule mining (ARM) is a data mining technique to discover interesting associations between datasets. The frequent pattern-growth (FP-growth) is an effective ARM algorithm for compressing information in the tree structure. However, it tends to suffer from the performance gap when processing large databases because of its mining procedure. This study presents a modified FP-growth (MFP-growth) algorithm to enhance the efficiency of the FP-growth by obviating the need for recurrent creation of conditional subtrees. The proposed algorithm uses a header table configuration to reduce the complexity of the whole frequent pattern tree. Four experimental series are conducted using different benchmark datasets to analyze the operating efficiency of the proposed MFP-growth algorithm compared with state-of-the-art machine learning algorithms in terms of runtime, memory consumption, and the effectiveness of generated rules. The experimental results confirm the superiority of the MFP-growth algorithm, which focuses on its potential implementations in various contexts.