Time Series Forecasting

Uni-FinLLM: A Unified Multimodal Large Language Model with Modular Task Heads for Micro-Level Stock Prediction and Macro-Level Systemic Risk Assessment ArXiv ID: 2601.02677 “View on arXiv” Authors: Gongao Zhang, Haijiang Zeng, Lu Jiang Abstract Financial institutions and regulators require systems that integrate heterogeneous data to assess risks from stock fluctuations to systemic vulnerabilities. Existing approaches often treat these tasks in isolation, failing to capture cross-scale dependencies. We propose Uni-FinLLM, a unified multimodal large language model that uses a shared Transformer backbone and modular task heads to jointly process financial text, numerical time series, fundamentals, and visual data. Through cross-modal attention and multi-task optimization, it learns a coherent representation for micro-, meso-, and macro-level predictions. Evaluated on stock forecasting, credit-risk assessment, and systemic-risk detection, Uni-FinLLM significantly outperforms baselines. It raises stock directional accuracy to 67.4% (from 61.7%), credit-risk accuracy to 84.1% (from 79.6%), and macro early-warning accuracy to 82.3%. Results validate that a unified multimodal LLM can jointly model asset behavior and systemic vulnerabilities, offering a scalable decision-support engine for finance. ...

Forward-Oriented Causal Observables for Non-Stationary Financial Markets ArXiv ID: 2512.24621 “View on arXiv” Authors: Lucas A. Souza Abstract We study short-horizon forecasting in financial time series under strict causal constraints, treating the market as a non-stationary stochastic system in which any predictive observable must be computable online from information available up to the decision time. Rather than proposing a machine-learning predictor or a direct price-forecast model, we focus on \emph{“constructing”} an interpretable causal signal from heterogeneous micro-features that encode complementary aspects of the dynamics (momentum, volume pressure, trend acceleration, and volatility-normalized price location). The construction combines (i) causal centering, (ii) linear aggregation into a composite observable, (iii) causal stabilization via a one-dimensional Kalman filter, and (iv) an adaptive ``forward-like’’ operator that mixes the composite signal with a smoothed causal derivative term. The resulting observable is mapped into a transparent decision functional and evaluated through realized cumulative returns and turnover. An application to high-frequency EURUSDT (1-minute) illustrates that causally constructed observables can exhibit substantial economic relevance in specific regimes, while degrading under subsequent regime shifts, highlighting both the potential and the limitations of causal signal design in non-stationary markets. ...

Quantitative Financial Modeling for Sri Lankan Markets: Approach Combining NLP, Clustering and Time-Series Forecasting ArXiv ID: 2512.20216 “View on arXiv” Authors: Linuk Perera Abstract This research introduces a novel quantitative methodology tailored for quantitative finance applications, enabling banks, stockbrokers, and investors to predict economic regimes and market signals in emerging markets, specifically Sri Lankan stock indices (S&P SL20 and ASPI) by integrating Environmental, Social, and Governance (ESG) sentiment analysis with macroeconomic indicators and advanced time-series forecasting. Designed to leverage quantitative techniques for enhanced risk assessment, portfolio optimization, and trading strategies in volatile environments, the architecture employs FinBERT, a transformer-based NLP model, to extract sentiment from ESG texts, followed by unsupervised clustering (UMAP/HDBSCAN) to identify 5 latent ESG regimes, validated via PCA. These regimes are mapped to economic conditions using a dense neural network and gradient boosting classifier, achieving 84.04% training and 82.0% validation accuracy. Concurrently, time-series models (SRNN, MLP, LSTM, GRU) forecast daily closing prices, with GRU attaining an R-squared of 0.801 and LSTM delivering 52.78% directional accuracy on intraday data. A strong correlation between S&P SL20 and S&P 500, observed through moving average and volatility trend plots, further bolsters forecasting precision. A rule-based fusion logic merges ESG and time-series outputs for final market signals. By addressing literature gaps that overlook emerging markets and holistic integration, this quant-driven framework combines global correlations and local sentiment analysis to offer scalable, accurate tools for quantitative finance professionals navigating complex markets like Sri Lanka. ...

Adaptive Weighted Genetic Algorithm-Optimized SVR for Robust Long-Term Forecasting of Global Stock Indices for investment decisions ArXiv ID: 2512.15113 “View on arXiv” Authors: Mohit Beniwal Abstract Long-term price forecasting remains a formidable challenge due to the inherent uncertainty over the long term, despite some success in short-term predictions. Nonetheless, accurate long-term forecasts are essential for high-net-worth individuals, institutional investors, and traders. The proposed improved genetic algorithm-optimized support vector regression (IGA-SVR) model is specifically designed for long-term price prediction of global indices. The performance of the IGA-SVR model is rigorously evaluated and compared against the state-of-the-art baseline models, the Long Short-Term Memory (LSTM), and the forward-validating genetic algorithm optimized support vector regression (OGA-SVR). Extensive testing was conducted on the five global indices, namely Nifty, Dow Jones Industrial Average (DJI), DAX Performance Index (DAX), Nikkei 225 (N225), and Shanghai Stock Exchange Composite Index (SSE) from 2021 to 2024 of daily price prediction up to a year. Overall, the proposed IGA-SVR model achieved a reduction in MAPE by 19.87% compared to LSTM and 50.03% compared to OGA-SVR, demonstrating its superior performance in long-term daily price forecasting of global indices. Further, the execution time for LSTM was approximately 20 times higher than that of IGA-SVR, highlighting the high accuracy and computational efficiency of the proposed model. The genetic algorithm selects the optimal hyperparameters of SVR by minimizing the arithmetic mean of the Mean Absolute Percentage Error (MAPE) calculated over the full training dataset and the most recent five years of training data. This purposefully designed training methodology adjusts for recent trends while retaining long-term trend information, thereby offering enhanced generalization compared to the LSTM and rolling-forward validation approach employed by OGA-SVR, which forgets long-term trends and suffers from recency bias. ...

EXFormer: A Multi-Scale Trend-Aware Transformer with Dynamic Variable Selection for Foreign Exchange Returns Prediction ArXiv ID: 2512.12727 “View on arXiv” Authors: Dinggao Liu, Robert Ślepaczuk, Zhenpeng Tang Abstract Accurately forecasting daily exchange rate returns represents a longstanding challenge in international finance, as the exchange rate returns are driven by a multitude of correlated market factors and exhibit high-frequency fluctuations. This paper proposes EXFormer, a novel Transformer-based architecture specifically designed for forecasting the daily exchange rate returns. We introduce a multi-scale trend-aware self-attention mechanism that employs parallel convolutional branches with differing receptive fields to align observations on the basis of local slopes, preserving long-range dependencies while remaining sensitive to regime shifts. A dynamic variable selector assigns time-varying importance weights to 28 exogenous covariates related to exchange rate returns, providing pre-hoc interpretability. An embedded squeeze-and-excitation block recalibrates channel responses to emphasize informative features and depress noise in the forecasting. Using the daily data for EUR/USD, USD/JPY, and GBP/USD, we conduct out-of-sample evaluations across five different sliding windows. EXFormer consistently outperforms the random walk and other baselines, improving directional accuracy by a statistically significant margin of up to 8.5–22.8%. In nearly one year of trading backtests, the model converts these gains into cumulative returns of 18%, 25%, and 18% for the three pairs, with Sharpe ratios exceeding 1.8. When conservative transaction costs and slippage are accounted for, EXFormer retains cumulative returns of 7%, 19%, and 9%, while other baselines achieve negative. The robustness checks further confirm the model’s superiority under high-volatility and bear-market regimes. EXFormer furnishes both economically valuable forecasts and transparent, time-varying insights into the drivers of exchange rate dynamics for international investors, corporations, and central bank practitioners. ...

Explainable Prediction of Economic Time Series Using IMFs and Neural Networks ArXiv ID: 2512.12499 “View on arXiv” Authors: Pablo Hidalgo, Julio E. Sandubete, Agustín García-García Abstract This study investigates the contribution of Intrinsic Mode Functions (IMFs) derived from economic time series to the predictive performance of neural network models, specifically Multilayer Perceptrons (MLP) and Long Short-Term Memory (LSTM) networks. To enhance interpretability, DeepSHAP is applied, which estimates the marginal contribution of each IMF while keeping the rest of the series intact. Results show that the last IMFs, representing long-term trends, are generally the most influential according to DeepSHAP, whereas high-frequency IMFs contribute less and may even introduce noise, as evidenced by improved metrics upon their removal. Differences between MLP and LSTM highlight the effect of model architecture on feature relevance distribution, with LSTM allocating importance more evenly across IMFs. ...

Bitcoin Forecasting with Classical Time Series Models on Prices and Volatility ArXiv ID: 2511.06224 “View on arXiv” Authors: Anmar Kareem, Alexander Aue Abstract This paper evaluates the performance of classical time series models in forecasting Bitcoin prices, focusing on ARIMA, SARIMA, GARCH, and EGARCH. Daily price data from 2010 to 2020 were analyzed, with models trained on the first 90 percent and tested on the final 10 percent. Forecast accuracy was assessed using MAE, RMSE, AIC, and BIC. The results show that ARIMA provided the strongest forecasts for short-run log-price dynamics, while EGARCH offered the best fit for volatility by capturing asymmetry in responses to shocks. These findings suggest that despite Bitcoin’s extreme volatility, classical time series models remain valuable for short-run forecasting. The study contributes to understanding cryptocurrency predictability and sets the stage for future work integrating machine learning and macroeconomic variables. ...

Reasoning on Time-Series for Financial Technical Analysis ArXiv ID: 2511.08616 “View on arXiv” Authors: Kelvin J. L. Koa, Jan Chen, Yunshan Ma, Huanhuan Zheng, Tat-Seng Chua Abstract While Large Language Models have been used to produce interpretable stock forecasts, they mainly focus on analyzing textual reports but not historical price data, also known as Technical Analysis. This task is challenging as it switches between domains: the stock price inputs and outputs lie in the time-series domain, while the reasoning step should be in natural language. In this work, we introduce Verbal Technical Analysis (VTA), a novel framework that combine verbal and latent reasoning to produce stock time-series forecasts that are both accurate and interpretable. To reason over time-series, we convert stock price data into textual annotations and optimize the reasoning trace using an inverse Mean Squared Error (MSE) reward objective. To produce time-series outputs from textual reasoning, we condition the outputs of a time-series backbone model on the reasoning-based attributes. Experiments on stock datasets across U.S., Chinese, and European markets show that VTA achieves state-of-the-art forecasting accuracy, while the reasoning traces also perform well on evaluation by industry experts. ...

Identifying and Quantifying Financial Bubbles with the Hyped Log-Periodic Power Law Model ArXiv ID: 2510.10878 “View on arXiv” Authors: Zheng Cao, Xingran Shao, Yuheng Yan, Helyette Geman Abstract We propose a novel model, the Hyped Log-Periodic Power Law Model (HLPPL), to the problem of quantifying and detecting financial bubbles, an ever-fascinating one for academics and practitioners alike. Bubble labels are generated using a Log-Periodic Power Law (LPPL) model, sentiment scores, and a hype index we introduced in previous research on NLP forecasting of stock return volatility. Using these tools, a dual-stream transformer model is trained with market data and machine learning methods, resulting in a time series of confidence scores as a Bubble Score. A distinctive feature of our framework is that it captures phases of extreme overpricing and underpricing within a unified structure. We achieve an average yield of 34.13 percentage annualized return when backtesting U.S. equities during the period 2018 to 2024, while the approach exhibits a remarkable generalization ability across industry sectors. Its conservative bias in predicting bubble periods minimizes false positives, a feature which is especially beneficial for market signaling and decision-making. Overall, this approach utilizes both theoretical and empirical advances for real-time positive and negative bubble identification and measurement with HLPPL signals. ...

From News to Returns: A Granger-Causal Hypergraph Transformer on the Sphere ArXiv ID: 2510.04357 “View on arXiv” Authors: Anoushka Harit, Zhongtian Sun, Jongmin Yu Abstract We propose the Causal Sphere Hypergraph Transformer (CSHT), a novel architecture for interpretable financial time-series forecasting that unifies \emph{“Granger-causal hypergraph structure”}, \emph{“Riemannian geometry”}, and \emph{“causally masked Transformer attention”}. CSHT models the directional influence of financial news and sentiment on asset returns by extracting multivariate Granger-causal dependencies, which are encoded as directional hyperedges on the surface of a hypersphere. Attention is constrained via angular masks that preserve both temporal directionality and geometric consistency. Evaluated on S&P 500 data from 2018 to 2023, including the 2020 COVID-19 shock, CSHT consistently outperforms baselines across return prediction, regime classification, and top-asset ranking tasks. By enforcing predictive causal structure and embedding variables in a Riemannian manifold, CSHT delivers both \emph{“robust generalisation across market regimes”} and \emph{“transparent attribution pathways”} from macroeconomic events to stock-level responses. These results suggest that CSHT is a principled and practical solution for trustworthy financial forecasting under uncertainty. ...