NDCG — the rank-aware metric for evaluating suggestion techniques
16 hours in the past
Suggestion techniques are all over the place. Because you’re studying this text, there’s an excellent likelihood Medium beneficial it in your feed. This text will discover NDCG — Normalized Discounted Cumulative Acquire, the rank-aware metric for evaluating any suggestion system mannequin.
Suggestion techniques assist customers uncover related objects like merchandise, profiles, posts, movies, adverts, or data primarily based on their preferences or conduct. These platforms deal with thousands and thousands of things, and displaying essentially the most related ones is vital to boosting consumer engagement and enterprise metrics. Corporations corresponding to Amazon, LinkedIn, Twitter, Instagram, Reddit, Spotify, YouTube, Netflix, Medium, and Quora use suggestion techniques of their apps.
These techniques are usually two-stage techniques consisting of a retrieval mannequin adopted by a rating mannequin. The retrieval mannequin funnels down essentially the most related objects from thousands and thousands of things primarily based on a similarity metric and passes them to the rating mannequin. The rating mannequin ranks the objects on a extra granular degree.