---
title: Multi-Horizon Stock Return Prediction via Transformer-Based Deep Learning
url: https://www.ml-quant.com/papers/ssrn/7605161/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-10-16
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: SSRN 7605161
source_url: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7605161
featured: 2026-10-16
citations: unknown
topic: ML & AI Methods
---


# Multi-Horizon Stock Return Prediction via Transformer-Based Deep Learning

Transformer-based models capture pricing signals via self-attention that boost out-of-sample Sharpe ratios across one-, three-, six-, and twelve-month stock return forecasts, outperforming standard machine learning approaches.

- Source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7605161
- Identifier: SSRN 7605161
- Released: 2026-10-14
- First featured: Quant Letter No. 135 (2026-10-16): https://www.ml-quant.com/issues/2026-10-16/
- Citations (Semantic Scholar): not tracked
- Published in: not yet
- Topic: ML & AI Methods
- Authors: Tobias Cramer, Wolfgang Drobetz, Tizian Otto

## Related

- [Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality](https://www.ml-quant.com/papers/arxiv/2405.21060/): The research identifies a link between state-space models and Transformers in deep learning, leading to the creation of a faster language modeling architecture, Mamba-2.
- [An Advanced Ensemble Deep Learning Framework for Stock Price Prediction Using VAE, Transformer, and LSTM Model](https://www.ml-quant.com/papers/arxiv/2503.22192/): The research introduces a combined deep learning framework for stock price prediction, demonstrating its effectiveness and reliability in predicting stock price movements.
- [Bytes Are All You Need: Transformers Operating Directly On File Bytes](https://www.ml-quant.com/papers/arxiv/2306.00238/): File Byte Transformers: ByteFormer is a deep learning model that enhances image classification accuracy by 5% and can perform audio classification without specific preprocessing, showcasing its versatility.
- [KiT: A Foundation Model for Financial Time-Series Forecasting using DiffusionTransformers](https://www.ml-quant.com/papers/arxiv/2609.34507/): Diffusion transformer foundation model reformulates candlestick prediction as conditional trajectory generation, achieving 0.057 mean return RankIC across markets and timescales.
- [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://www.ml-quant.com/papers/arxiv/2312.00752/): Sequence Modeling: Mamba, a neural network architecture that doesn't use attention or MLP blocks, provides faster inference and better performance in language, audio, and genomics than Transformers.
- [Octo: An Open-Source Generalist Robot Policy](https://www.ml-quant.com/papers/arxiv/2405.12213/): Octo is a large transformer-based policy for robotic manipulation, trained on a vast dataset, that can be instructed via language or images and adapted to new domains.