Full Program »
Performance Comparisons of Association Rule Learning Algorithms On A Supermarket Data
In today's business environment, sustaining growth and profitability is never a guarantee. This concept of instability encourages business owners to seek new opportunities using technological and scientific advances to grasp the sustainability and maximizing profit. Nowadays, Machine Learning is helping the retail sector in many ways. Such improved life cycles of products, services, and business models. Association Rule Learning (ARL) is an unsupervised machine learning technique having great potential to extract interesting and useful associations and/or correlations among frequent item sets in transactional dataset. ARL techniques are widely employed in a variety of applications, including web click data, medical diagnostics, consumer behavior analysis, bioinformatics, and many more. In this paper, we conduct a comprehensive performance analysis of six distinct ARL algorithms, which was carried out to find the best fit algorithm for our data. The data is a large transactional dataset of one of the biggest supermarket chains in Saudi Arabia. Those algorithms are; Apriori, FP-Growth, GCD, Top-K Rules, TNR, and Closed AR. Note that these algorithms are uniquely different in terms of the number of datasets scanning which influences performance. The evaluation study shows that FP-growth algorithm based on FP-Tree outperformed other algorithms. FP-growth was more efficient and scalable than other algorithms for the given data in terms of memory usage and processing time.