Association Rule Mining on Noisy Synthetic Data: Challenges and Solutions

A fundamental data mining technique used to uncover hidden patterns in large datasets is known as Association Rule Mining (ARM). However, the presence of noise in data significantly impacts the accuracy and reliability of extracted rules. This paper explores the challenges of mining association rules in noisy synthetic datasets and evaluates the effects of different noise levels on rule extraction. A synthetic dataset of 50,000 transactions was generated, incorporating controlled noise to simulate real-world inconsistencies. The study applies the Apriori and FP-Growth algorithms to analyze how noise affects rule generation and proposes preprocessing techniques to mitigate its negative impact. Experimental results demonstrate that noise reduces rule confidence and support, leading to the suppression of meaningful patterns and the generation of misleading associations. The proposed preprocessing strategies improve rule accuracy, highlighting the importance of noise handling in ARM. This research contributes to the development of robust, noise-resistant mining techniques to enhance data-driven decision-making in noisy environments.