<![CDATA[Association Rule Mining (ARM) is one of unsupervised learning approach of machine learning. It acts as a data analysis technique that enables the identification of frequent patterns, correlations, associations, and causal structures within certain datasets. This method widely used in numerous studies and practices to explore knowledges and strengthen decision making. However, dealing a large dataset with high number of transactions may become the shortcoming for the ARM algorithms, such as Apriori, FP-Growth, and Eclat. It leads them to face several challenges, including computational complexity, long mining durations, and memory consumption. Hence, this paper proposes k-means clustering to generates several groups of data to handle the issue, then proceed the ARM algorithms for each individual produced cluster. The study used Elbow method and Silhouette Coefficient as the method to determining optimum number of clusters to be used. The result pointed out that k-means-ARM generates a greater number of rules and provides more contextually relevant and significant correlations. In term of Lift Ratio average score, the k-means-ARM shows the greater value rather than non k-means ARM. The k-means-ARM combination is robust; this approach improves the efficiency and scalability of ARM for large datasets and enhances the interpretability of the discovered association rules]]>
