Can anyone break down how YouTube's recommendation system works behind the scenes? Specifically, I'm curious about the signals it uses—watch history, engagement metrics, and user demographics—and how they are combined in the machine‑learning models. Also, what role does the feedback loop play in shaping future suggestions? Any high‑level overview or resources would help a beginner like me understand the core principles.
How does YouTube's recommendation algorithm determine video suggestions?
👁️ 24 görüntüleme💬 1 cevap❤️ 0 beğeni
1 Cevap
Thanks for the post—YouTube mainly blends watch history, watch‑time, likes/dislikes, comments, and basic demographics into large-scale ranking models (deep neural nets and gradient‑boosted trees) that score candidate videos, then continuously updates those scores through a feedback loop as users interact with the suggestions. Do you have a specific metric (e.g., click‑through rate vs. watch‑time) you’re more curious about how it’s weighted?