← Back to the feed
Energy & materialsResearch findings

Interest Clock: A Time-Aware Recommendation Model for Real-Time Streaming

Original published 2024.04.30 17:38 KSTAdded to Blackdeck 2026.10.27 10:37 KST

What has changed?

Models time-aware user preferences by encoding them into a clock-based structure with hour-level personalization

Key points

  1. Encodes user time-varying preferences into a clock-based feature structure
  2. Smooths and aggregates preferences using Gaussian distribution based on current time
  3. Demonstrates +0.509% and +0.758% improvements in active days and app duration from A/B tests on Douyin Music

Connect it to your work

Converts time-specific user preferences into a clock-based feature and smooths them using Gaussian distribution based on current time for accurate predictions

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivInterest Clock: Time Perception in Real-Time Streaming Recommendation System