Skip to content
Book Open access

ZoomLens Stock Forecasting: Adaptive Multi-Scale Modeling for Market-Aware Investment

Nov 2026 · Proceedings of the 35th ACM International Conference on Information and Knowledge Management · 0 citations · 38 references

Abstract

Modern stock markets demand a sophisticated capability to simultaneously focus on fine-grained local volatility and perceive broader global trends. However, existing methods often remain limited by rigid temporal architectures and a neglect of dynamic cross-asset dependencies. To address these challenges, we propose an Adaptive Multi-scale signal Selection and Fusion Model (AMSFM). Functioning as an intelligent ZoomLens, AMSFM incorporates an adaptive temporal scale selector that dynamically aligns modeling horizons with real-time market regimes, effectively capturing both microscopic fluctuations and macroscopic contexts. Furthermore, we propose an interpretable attribution and interaction mechanism to decompose forecasts into transparent feature contributions. To capture broader market influences, our model employs a market-aware forecasting strategy that integrates global context to perceive panoramic spillover effects without relying on rigid prior knowledge. Extensive experiments on three real-world datasets demonstrate that AMSFM significantly outperforms state-of-the-art benchmarks. The source code is available at: https://github.com/yizhibeijibei/AMSFM.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.