Anchoring Trends: Mitigating Social Media Popularity Prediction Drift via Feature Clustering and Expansion

TL;DR

This paper introduces a novel framework that uses feature clustering and semantic anchors generated by large language models to improve the robustness of social media popularity predictions against trend shifts.

Contribution

It proposes the AMCFG framework that discovers invariant patterns and generates semantic anchor features to mitigate prediction drift in video popularity forecasting.

Findings

Significantly improves prediction accuracy on out-of-distribution data.

Enhances temporal robustness of popularity prediction models.

Demonstrates effectiveness across multiple social media datasets.

Abstract

Predicting online video popularity faces a critical challenge: prediction drift, where models trained on historical data rapidly degrade due to evolving viral trends and user behaviors. To address this temporal distribution shift, we propose an Anchored Multi-modal Clustering and Feature Generation (AMCFG) framework that discovers temporally-invariant patterns across data distributions. Our approach employs multi-modal clustering to reveal content structure, then leverages Large Language Models (LLMs) to generate semantic Anchor Features, such as audience demographics, content themes, and engagement patterns that transcend superficial trend variations. These semantic anchors, combined with cluster-derived statistical features, enable prediction based on stable principles rather than ephemeral signals. Experiments demonstrate that AMCFG significantly enhances both predictive accuracy and temporal robustness, achieving superior performance on out-of-distribution data and providing a viable solution for real-world video popularity prediction.