Data-Driven Insights for Enhancing Retail Sustainability: An Empirical Dimension Reduction Apriori Algorithm for Association Rule Mining

In the era of digital transformation and big data, organizations generate vast volumes of raw data from diverse channels such as IoT devices, cyber systems, and e-commerce platforms. Extracting meaningful insights from this data is essential, particularly in identifying interesting item sets through association rule mining. These patterns reveal strong product relationships and can drive more sustainable strategies for marketing and sales in retail and online platforms. However, when working with large-scale datasets containing thousands of transactions and items, many frequently occurring items do not necessarily result in meaningful or interesting item sets. This leads to inefficiencies in the mining process and unnecessary computational overhead. To address this issue, we propose a novel algorithm called Dimension Reduction Apriori (DR-Apriori), which enhances the performance of traditional association rule mining by incorporating a dimension reduction step. This approach streamlines the dataset, enabling the extraction of more relevant and interesting item sets. We evaluated DR-Apriori on two benchmark datasets—15,000 transactions with 4,089 features, and 7,500 transactions with 121 features—under varying thresholds for dimension reduction, support, and confidence. Experimental results show that DR-Apriori outperforms traditional Apriori and Hybrid-Apriori algorithms, achieving up to 51% faster runtime, reducing memory usage by up to 31%, and maintaining the number of interesting item sets discovered. This study highlights the potential of DR-Apriori in enhancing the efficiency and scalability of association rule mining, ultimately supporting more intelligent, data-driven decisions for sustainable retail practices.