Towards More Satisfactory Recommendations: Leveraging Bias Awareness in Recommender Systems

[ma14ch]

I am working on a recommendation system, and despite improving various algorithms and incorporating user data, I have realized the need for personalization to enhance user satisfaction. This personalization entails introducing a numerical bias parameter that is injected into the suggested results for each user. The question I am grappling with is: what is the best method for injecting this bias, and how can we evaluate its accuracy in a metric-based manner?