Stock Return Prediction Based on Financial Text Sentiment Analysis and Dynamic Lead-Lag Relationships in FF5 Residual Networks

DOI

10.2991/978-94-6239-830-6_15How to use a DOI?

Keywords

stock return prediction; financial text sentiment; FF5 residuals; Lead-Lag network; Missing Link

Abstract

Stock return prediction remains challenging because public information, investor attention and cross-stock signals are incorporated into prices at different speeds. This study develops a two-layer empirical framework for a fixed universe of 150 U.S.-listed stocks. The prediction layer first estimates positive-forward-return probabilities with a PatchTST-based price model and then combines them with lagged sentiment features from investor comments, financial news and formal disclosures to measure the incremental contribution of text. Separately, the relationship layer removes common factor-driven co-movements with FF5 residuals and constructs dynamic Lead-Lag networks to identify Missing Link opportunities in delayed cross-stock reactions. The contribution of text varies across sources and horizons: investor-comment information is more useful at short and short-to-medium horizons, whereas comments combined with formal disclosures are more stable at medium and long horizons. The strongest representative Missing Link results occur in 60–70-day graph windows with 5–7-day follower periods, while the broader short-follower-period grid shows that positive evidence is concentrated in 40–70-day dynamic graph windows.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Juntao Zhou
PY  - 2026
DA  - 2026/11/14
TI  - Stock Return Prediction Based on Financial Text Sentiment Analysis and Dynamic Lead-Lag Relationships in FF5 Residual Networks
BT  - Proceedings of the 6th International Conference on Internet Finance and Digital Economy (ICIFDE 2026)
PB  - Atlantis Press
SP  - 182
EP  - 191
SN  - 2352-5428
UR  - https://doi.org/10.2991/978-94-6239-830-6_15
DO  - 10.2991/978-94-6239-830-6_15
ID  - Zhou2026
ER  -