Paper: arXiv 2503.06928
Authors: Yanlong Wang, Jian Xu, Tiantian Gao, Hongkang Zhang, Shao-Lun Huang, Danny Dongning Sun, Xiao-Ping Zhang
Abstract
Despite the growing attention to time series forecasting in recent years, many studies have proposed various solutions to address the challenges encountered in time series prediction, aiming to improve forecasting performance. However, effectively applying these time series forecasting models to the field of financial asset pricing remains a challenging issue. There is still a need for a bridge to connect cutting-edge time series forecasting models with financial asset pricing. To bridge this gap, we have undertaken the following efforts: 1) We constructed three datasets from the financial domain; 2) We selected over ten time series forecasting models from recent studies and validated their performance in financial time series; 3) We developed new metrics, msIC and msIR, in addition to MSE and MAE, to showcase the time series correlation captured by the models; 4) We designed financial-specific tasks for these three datasets and assessed the practical performance and application potential of these forecasting models in important financial problems. We hope the developed new evaluation suite, FinTSBridge, can provide valuable insights into the effectiveness and robustness of advanced forecasting models in finanical domains.
Complexity vs Empirical Score
- Math Complexity: 4.0/10
- Empirical Rigor: 6.0/10 (capped from 8.5: transaction costs not modelled; see penalty flags)
- Quadrant: Street Traders — empirical and applied, lower math complexity (effective empirical rigor capped from 8.5)
Why this score: The paper’s core contribution is an empirical evaluation suite with curated datasets, specific financial tasks, and new metrics (msIC, msIR), demonstrating high practical rigor. While it references advanced time series models (e.g., transformers, decomposition methods), the paper itself is more focused on application and benchmarking rather than deriving new mathematical theorems.
Research Flowchart
flowchart TD
A["Research Goal: Bridge TS models with Financial Asset Pricing"] --> B{"Methodology"}
B --> C["Data Creation<br>3 Financial Datasets"]
B --> D["Model Selection<br>10+ TS Forecasting Models"]
B --> E["Metric Development<br>msIC, msIR, MSE, MAE"]
C --> F["Financial Task Design"]
D --> F
E --> F
F --> G["Computational Process:<br>Benchmarking & Evaluation"]
G --> H["Key Findings:<br>FinTSBridge Evaluation Suite<br>Model Insights & Practical Potential"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper introduces FinTSBridge, a suite that provides three financial time series datasets, selects over ten forecasting models, and adds new multi‑step correlation metrics msIC and msIR for evaluating financial forecasting.
Source sentence
We developed new metrics, msIC and msIR, in addition to MSE and MAE, to showcase the time series correlation captured by the models
Assumes 2
- The investment strategy rebalances every five trading days. execution
Source sentence
The investment strategy involves rebalancing every five trading days by selecting from the 20 available indices
- Risk‑free rate is proxied by the 1‑Year Shibor. data
Source sentence
Risk-free rate (proxied by 1-Year Shibor)
Data
SPX index · equities · daily
- GSMI market-prices public
Source sentence
The GSMI dataset includes 20 major indices from global stock markets, recording daily price and trading volume data for these indices over nearly 20 years, from 2005 to 2024
- OPTION options public
Source sentence
The OPTION dataset includes the CSI 300ETF options from the Chinese financial market, with variables related to risk for both call and put options
- BTCF crypto-onchain public
Source sentence
The BTCF dataset is sourced from the Binance platform and contains hourly frequency data for Bitcoin spot and futures contracts, spanning from January 1, 2020, to November 30, 2024
Costs
Not modeled
Finds 2
Sharpe 1.20 (benchmark 0.02) · annual return 17.9% · total return 133.7% · max drawdown 21.6%
- The Naive model performs poorly on strongly periodic series like Traffic but yields smaller errors on less periodic series such as Exchange, indicating prediction error alone is insufficient for evaluation. predictive-accuracy
Source sentence
The results suggest that the Naive model performs poorly on time series with strong periodicity, such as Traffic, but achieves smaller prediction errors on datasets like Exchange
- Extensive strategy evaluations and visualizations validate the effectiveness and potential of advanced models in real‑world financial deployment. performance
Source sentence
extensive strategy evaluations and visualizations validate the effectiveness and potential of advanced models in real-world financial deployment
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 11 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (61)
- Dynamic connectedness between stock markets in the presence of the COVID-19 pandemic: does economic policy uncertainty matter? (2021)
- Indexing and stock market serial dependence around the world (2019)
- How do crises spread? Evidence from accessible and inaccessible stock indices (2006)
- Time-varying pattern causality inference in global stock markets (2021)
- Efficient capital markets: A review of theory and empirical work (1970)
- Stock market prices do not follow random walks: Evidence from a simple specification test (1988)
- Returns to Buying Winners and Selling Losers: Implications for (1993)
- Multifactor explanations of asset pricing anomalies (1996)
and 53 more
- The efficient market hypothesis and its critics (2003)
- Machine learning and the stock market (2023)
- Machine learning in the Chinese stock market (2022)
- Empirical asset pricing via machine learning (2020)
- Taming the factor zoo: A test of new factors (2020)
- Comovement (2005)
- TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis (2023) — link
- Timexer: Empowering transformers for time series forecasting with exogenous variables (2024) — link
- TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (2024)
- A Time Series is Worth 64 Words: Long-term Forecasting with Transformers (2023)
- Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting (2021)
- Informer: Beyond efficient transformer for long sequence time-series forecasting (2021)
- Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting (2022)
- Crossformer: Transformer Utilizing Cross-Dimension Dependency for Multivariate Time Series Forecasting (2023) — link
- iTransformer : Inverted Transformers Are Effective for Time Series Forecasting (2024) — link
- TSM ixer: An All- MLP Architecture for Time Series Forecast-ing (2023) — link
- N-BEATS: Neural basis expansion analysis for interpretable time series forecasting (2020) — link
- Are transformers effective for time series forecasting? (2023)
- Non-stationary transformers: Exploring the stationarity in time series forecasting (2022)
- Koopa: Learning non-stationary time series dynamics with koopman predictors (2023)
- Micn: Multi-scale local and global context modeling for long-term series forecasting (2023)
- Spectral temporal graph neural network for multivariate time-series forecasting (2020)
- Reversible instance normalization for accurate time-series forecasting against distribution shift (2021)
- Long-term Forecasting with Ti DE : Time-series Dense Encoder (2023)
- N-BEATS: Neural basis expansion analysis for interpretable time series forecasting (2019)
- Deep learning for forecasting stock returns in the cross-section (2018)
- Decision support from financial disclosures with deep neural networks and transfer learning (2017)
- Deep learning for event-driven stock prediction (2015)
- Mastering Stock Markets with Efficient Mixture of Diversified Trading Experts (2023) — link
- AlphaFin: Benchmarking Financial Analysis with Retrieval-Augmented Stock-Chain Framework (2024)
- A Survey of Large Language Models for Financial Applications: Progress, Prospects and Challenges — → 2406.11903
- MacroHFT: Memory Augmented Context-aware Reinforcement Learning On High Frequency Trading — → 2406.14537
- Understand the current status, trends and challenges of cryptocurrency derivatives (2020)
- Relations among Bitcoin Futures, Bitcoin Spot, Investor Attention, and Sentiment (2023)
- Attention is all you need (2017)
- PSformer: Parameter-efficient Transformer with Segment Attention for Time Series Forecasting (2024) — link
- Deep time series models: A comprehensive survey and benchmark (2024) — link
- Adaptive Normalization for Non-stationary Time Series Forecasting: A Temporal Slice Perspective (2023)
- LLMs and Foundational Models: Not (Yet) as Good as Hoped. (2024)
- Commentary: Can LLMs Provide Good Forecasts? (2024)
- Are Language Models Actually Useful for Time Series Forecasting? (2024) — link
- Fundamental limitations of foundational forecasting models: The need for multimodality and rigorous evaluation (2024) — link
- How to use security analysis to improve portfolio selection (1973)
- Active Portfolio Management: A Quantitative Approach for Producing Superior Returns and Controlling Risk (2000)
- A survey on machine learning models for financial time series forecasting (2022)
- Time-series forecasting with deep learning: a survey (2021)
- Machine learning advances for time series forecasting (2023)
- Forecasting stock market volatility with a large number of predictors: New evidence from the MS-MIDAS-LASSO model (2022)
- The risk and return of equity and credit index options (2024)
- Risk spillovers across geopolitical risk and global financial markets (2023)
- Advances in financial machine learning (2018)
- Automatic de-biased temporal-relational modeling for stock investment recommendation (2024)
- Data-driven stock forecasting models based on neural networks: A review (2024)
2 of 61 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.