You are building a recommendation system for a digital platform with personalized feeds and item suggestions. The system works well for active users and established items, but it struggles when a new user arrives or a new item is added with little or no interaction history.
How would you handle cold start for a new user or item in a recommendation system?