<![CDATA[Background: Many studies have compared the Apriori, FP-Growth, and ECLAT algorithms. Most of the previous literature focuses on runtime evaluation (execution speed) on homogeneous datasets. Research specifically mapping the visualization of memory curves against the level of data sparsity on common e-commerce heterogeneous. Aims: This study aims to analyze the relationship between sparsity and memory usage of three classic Association Rule Mining algorithms such as Apriori, FP-Growth and ECLAT using a heterogeneous e-commerce dataset. Method: A quantitative approach using Association Rule Mining was applied to different raw datasets from 721 to 30000 transactions. Transactions focused on various item types. The Apriori, FP-Growth, and ECLAT algorithms are implemented in Python, using grid search for minimum support and confidence adjustment to generate frequent item sets and association rules for performance comparison. Result: The results show that the relationship between sparsity and memory usage differs significantly in the three classical association rule mining algorithms. The ECLAT algorithm shows low memory consumption because it uses a vertical TID-list, while the Apriori and FP-Growth algorithms show high memory consumption because they are related to their pattern search methods (Candidate Generation and FP-Tree). Conclusion: This study shows that memory usage is not only influenced by sparsity, but also by the characteristics of the dataset. Then the results of this analysis provide an overview of heterogeneous e-commerce datasets when used in the classical association rule mining algorithm.]]>
Copyrights © 2026
