International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 02 | Feb 2024
p-ISSN: 2395-0072
www.irjet.net
MODEL BIAS IN RECOMMENDATION SYSTEMS: UNDERSTANDING, IMPACT, AND MITIGATION TECHNIQUES Vatesh Pasrija Meta Platforms Inc, Seattle, USA. ---------------------------------------------------------------------------***--------------------------------------------------------------------------I. INTRODUCTION TO RECOMMENDATION SYSTEM Recommendation systems are extensively employed in several domains, including e-commerce, content streaming platforms, social media, and customized news providers. These systems strive to offer consumers personalized recommendations by analyzing their preferences, behavioral history, and other related information. However, an essential aspect that requires attention in recommendation systems is the presence of model bias. Model bias refers to the inherent biases that may exist in recommendation systems, resulting in unequal treatment or unfair advantages for specific items or individuals. Popularity bias is a c ommon form of model bias in recommendation systems. Popularity bias pertains to the inclination of recommendation algorithms to prioritize popular products over less popular ones [1]. This bias arises due to the fact that popular items tend to have a greater quantity of ratings and interactions, hence increasing the likelihood of them being recommended to users [2]. The presence of popularity bias can lead to adverse effects, as it might result in excessive focus on popular items, while ignoring niche or long-tail items that may be of relevance to particular users. To address the issue of popularity bias in recommendation systems, various methodologies have been suggested in academic literature.The most common strategies to fix the model bias in recommendation systems are to include a wider range of items in the suggestions, use fairness-aware algorithms that try to give all users the same recommendations, and add personalization methods that take into account each person's likes, dislikes, and relevant information. Understanding and resolving model bias in recommendation systems is essential to guarantee equal and precise recommendations, enhance user satisfaction, and promote diversity and serendipity in the recommendation process. As we begin to investigate the complexities of model bias in recommendation systems, Figure 1 provides a fundamental understanding of this phenomenon. Keywords: Recommendation systems, Model bias, Popularity bias, Fairness-aware algorithms, Academic literature
II.
UNDERSTANDING MODEL BIAS IN RECOMMENDATION SYSTEMS
Understanding model bias in recommendation systems is acknowledging the inherent biases that may exist and influence the recommendations given to users. This involves recognizing biases such as popularity bias, which occurs when more popular things are given preference over less popular ones [3]. Through understanding of model bias, recommendation systems can be designed and enhanced to deliver equitable and impartial suggestions to all users, irrespective of their preferences or the popularity of the suggested goods. The influence of model bias in recommendation systems can have extensive and detrimental effects. Popularity bias can result in a lack of diversity in the suggested items,
© 2024, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 452