Untangling the Role of Data Processing in Enhancing Recommender Systems Performance

Abstract

Recommender systems, pivotal in guiding user choices in numerous online platforms, rely heavily on the integrity and quality of the underlying data. This study delves into the transformative potential of data preprocessing on the effectiveness of such systems, using the Sephora cosmetics dataset as a case study. By juxtaposing results from both Collaborative Filtering (CF) and Content-Based Filtering (CBF) approaches—with and without preprocessing—we unveil the marked improvements in prediction accuracy post-preprocessing. Our findings accentuate that meticulous data preparation isn't merely a precursor but a fundamental aspect of optimizing recommender systems. Through this exploration, we aim to highlight best practices and underscore the inextricable link between data hygiene and the success of recommendation algorithms.