Long-Term Interest Clock: Fine-Grained Time Perception in Streaming Recommendation System

2025-01-27

AI Summary

Existing recommender systems struggle to model fine-grained intra-day temporal evolution of user interests (e.g., at hourly or half-hourly resolution), limiting recommendation accuracy in time-sensitive scenarios such as commuting, lunch breaks, and bedtime. To address this, we propose Clock-Rec—a streaming recommendation framework featuring a novel long-short-term behavioral co-modeling mechanism with fine-grained time awareness. Clock-Rec introduces two core modules: (i) Clock-GSU, which enables candidate-item-driven real-time subsequence retrieval; and (ii) Clock-ESU, which incorporates slot-aware attention and dynamic time-weighted aggregation to enable context-aware, moment-specific modeling of long-term behaviors. Unlike conventional approaches relying on coarse hourly discrete embeddings or static interest clocks, Clock-Rec dynamically focuses long-term behavior representations conditioned on the current temporal context. Deployed on the Douyin Music app, A/B testing demonstrates a 0.122% increase in user active days and significant improvements in recommendation precision during critical intra-day periods—including midday, commuting, and pre-sleep hours.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Recommender Systems
• Planning, Routing, and Scheduling: Activity and Plan Recognition

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Web Mining and Content Analysis: Content-based information diffusion
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

User interests manifest a dynamic pattern within the course of a day, e.g., a user usually favors soft music at 8 a.m. but may turn to ambient music at 10 p.m. To model dynamic interests in a day, hour embedding is widely used in traditional daily-trained industrial recommendation systems. However, its discreteness can cause periodical online patterns and instability in recent streaming recommendation systems. Recently, Interest Clock has achieved remarkable performance in streaming recommendation systems. Nevertheless, it models users' dynamic interests in a coarse-grained manner, merely encoding users' discrete interests of 24 hours from short-term behaviors. In this paper, we propose a fine-grained method for perceiving time information for streaming recommendation systems, named Long-term Interest Clock (LIC). The key idea of LIC is adaptively calculating current user interests by taking into consideration the relevance of long-term behaviors around current time (e.g., 8 a.m.) given a candidate item. LIC consists of two modules: (1) Clock-GSU retrieves a sub-sequence by searching through long-term behaviors, using query information from a candidate item and current time, (2) Clock-ESU employs a time-gap-aware attention mechanism to aggregate sub-sequence with the candidate item. With Clock-GSU and Clock-ESU, LIC is capable of capturing users' dynamic fine-grained interests from long-term behaviors. We conduct online A/B tests, obtaining +0.122% improvements on user active days. Besides, the extended offline experiments show improvements as well. Long-term Interest Clock has been integrated into Douyin Music App's recommendation system.

Problem

• Temporal Preferences
• Recommendation Systems
• User Engagement

Innovation

• Long-term Interest Clock (LIC)
• Temporal Preference Modeling
• Accurate Recommendation Prediction