A NOVEL ALGORITHM FOR EFFICIENT FREQUENT PATTERN MINING

Abstract

The rapid expansion of digital data across various sectors—including retail, healthcare, education, and finance—has increased the need for effective techniques that can extract meaningful knowledge from large databases. Among these techniques, association rule mining plays a crucial role in identifying frequent itemsets and uncovering hidden relationships within transactional data. Although traditional algorithms such as Apriori, FP-Growth, and Eclat are widely used for this purpose, their performance depends heavily on a user-defined minimum support threshold. Selecting this value without proper domain understanding often leads to inappropriate outputs, either by generating a large number of insignificant patterns or by overlooking valuable associations.To overcome these limitations, this research introduces a new association rule–based mining algorithm that determines the minimum support threshold mathematically instead of relying on user input. This automated approach simplifies the mining process, enhances accuracy, and reduces the need for expert intervention. The algorithm is implemented using Python and tested on standard datasets obtained from the UCI Repository. Its effectiveness is evaluated by comparing the number of frequent itemsets generated and the execution time with those of the conventional Apriori algorithm. The experimental analysis reveals that the proposed method consistently outperforms Apriori. It produces fewer redundant frequent itemsets, resulting in lower memory usage and improved clarity in the discovered patterns. Additionally, the algorithm demonstrates faster execution times across datasets of varying sizes, highlighting its efficiency and scalability. By eliminating manual threshold selection and improving computational performance, the proposed approach offers a more reliable, practical, and user-friendly solution for frequent pattern mining. This contributes to more intelligent and automated decision-support systems capable of handling large and complex datasets.