Journal of Forecasting最新文献

{"title":"An Econometric Framework to Nowcast Low-Frequency Data","authors":"Irfan A. Qureshi,&nbsp;Arief Ramayandi,&nbsp;Ghufran Ahmad","doi":"10.1002/for.70102","DOIUrl":"https://doi.org/10.1002/for.70102","url":null,"abstract":"<div>\u0000 \u0000 <p>Standard nowcasting frameworks commonly use weekly or monthly variables to monitor quarterly gross domestic product (GDP). However, this method is not suitable for economies that track GDP annually. We modify the state-space representation of an otherwise standard dynamic factor model to represent annual variables as a linear combination of latent monthly indicators for more frequently released variables. Using data from a lower middle-income country, we derive a monthly activity measure that effectively tracks annual GDP growth. These estimates outperform institutional forecasts and competing approaches to estimate low-frequency data. The model offers broader applications to countries facing data limitations, especially lower income countries.</p>\u0000 </div>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1473-1494"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148147021","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Neural Network Particle Filter for Time Series Data","authors":"Dewi E. W. Peerlings,&nbsp;Jan A. van den Brakel,&nbsp;Nalan Baştürk","doi":"10.1002/for.70100","DOIUrl":"https://doi.org/10.1002/for.70100","url":null,"abstract":"<div>\u0000 \u0000 <p>For the analysis of nonlinear or non-Gaussian state space models (SSMs), extended Kalman filters and particle filters (PFs) are proposed in the literature. Although these filters allow to formulate SSMs that are much more flexible compared to the linear Gaussian model, they are still based on parametric distributions. In this paper, a novel PF is proposed for the analysis of high-frequency time series with heavy tails and outliers such as GPS data, road sensor data, climate data, social media data, and data on stock prices. A neural network (NN) is trained using multiple time series to obtain a nonparametric approximation of the probability density function for the observation equation of the SSM and is combined with a PF to obtain estimates for the unobserved states of a local level model. This results in the so-called neural network particle filter (NNPF). We illustrate the accuracy gains from our proposed method in an extended simulation where time series are generated under the assumption of a local level model and Gaussian, Student's \u0000<span></span><math>\u0000 <mi>t</mi></math>, noncentral Student's \u0000<span></span><math>\u0000 <mi>t</mi></math>, and Poisson distributions for the observation equation. The proposed NNPF outperforms existing filters, particularly in the case of continuous distributions with heavy tails and outliers. The proposed NNPF is applied to a real-life application using vehicle minute counts obtained with road sensors.</p>\u0000 </div>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1438-1453"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148147210","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"On Capturing Multi-Scale Market Dynamics for High-Frequency Stock Price Forecasting Using a Hybrid Attention-Based Deep Learning Model","authors":"Runze Jiang,&nbsp;Yuping Song","doi":"10.1002/for.70094","DOIUrl":"https://doi.org/10.1002/for.70094","url":null,"abstract":"<div>\u0000 \u0000 <p>Forecasting high-frequency stock prices is a significant challenge due to inherent noise, non-stationarity, and complex market dynamics. Conventional models often struggle to effectively extract meaningful signals from raw price and volume data. To address this, we propose an innovative hybrid framework, EWOA-VMD-ATT-BiGRU, which introduces several key novelties to enhance prediction accuracy. First, our model uniquely decomposes both price and volume sequences as an innovative feature engineering, which allows for the unveiling of multi-scale market characteristics and effectively mitigates signal interference. Second, we employ an enhanced whale optimization algorithm (EWOA) to adaptively optimize the variational mode decomposition (VMD) parameters, ensuring a more precise and data-driven signal separation. Finally, a Bidirectional GRU network integrated with an Attention mechanism (ATT-BiGRU) is utilized to dynamically weigh the importance of the decomposed features for superior prediction. Empirical results on the high-frequency SSE index demonstrate our model's superior performance. Compared to the second-ranked model, it achieves reductions in MSE, RMSE, MAE, MSLE, MAPE, and SMAPE by 6.23%, 3.17%, 3.22%, 5.87%, 3.16%, and 3.15%, respectively. Notably, our strategy of decomposing both price and volume yields a substantial improvement, reducing key error metrics by over 40% compared to an equivalent non-decomposed model. Our proposed decomposition-based hybrid model offers a more precise and robust approach for forecasting of high-frequency stock prices.</p>\u0000 </div>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1399-1420"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148146971","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Dimensionality-Aware Credit Scoring With Hybrid Feature Selection and Simplified Graph Convolutional Networks","authors":"Anusha Hegde,&nbsp;Biswajit Bhowmik","doi":"10.1002/for.70092","DOIUrl":"https://doi.org/10.1002/for.70092","url":null,"abstract":"<div>\u0000 \u0000 <p>Peer-to-peer (P2P) lending is an emerging Financial Technology (FinTech) innovation that allows direct loans between individuals and businesses, avoiding traditional intermediaries. Contemporarily, loan defaults become a significant risk for financial institutions, which prompts the need for better default risk prediction models. Literature includes salient issues including feature selection, separation of outliers, and the effective integration of graph-based neural networks towards enhancing such a prediction model. This work proposes a novel hybrid feature selection method combining Boruta with minimum Redundancy Maximum Relevance (mRMR) for an enhanced credit default prediction. The selected feature set is then tested against tree-based models, which are particularly effective because they can model nonlinear relationships and interactions between variables. However, tree-based models cannot capture complex dependencies very well and fail to benefit from inherent structural relationships in data. Next, we develop a new approach using k-Nearest Neighbors (kNN), GraphSAGE and, Simplified Graph Convolution Networks (SGCN). In addition, we use Isolation Forest, an unsupervised anomaly detection method, to distinguish clean and noisy data so that the proposed model can be grounded in credible and relevant subsets of data. The experimentation is performed on the complete Lending Club and Prosper Datasets (including their inlier and outlier subsets). The findings using the Lending Club Dataset highlight the hybrid feature selection method as an effective and robust strategy for improving SGCN performance, especially in the presence of diverse and noisy data. Comparison study shows that the proposed approach improves the metrics such as accuracy, F1-score, and AUC, upto 88.99%, 88.61%, and 96.37%, respectively. It ensures that the proposed approach outperforms existing many approaches on multiple subsets of data. It enhances the robustness of the proposel model and suited very well for credit default prediction in P2P lending.</p>\u0000 </div>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1368-1398"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148147103","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Forecasting Financial Risk With Minute-Level Transaction Data and Economic Policy Uncertainty: A New Mixed-Frequency Time-Varying Forecasting Framework","authors":"Shuai Wang,&nbsp;Jianzhou Wang,&nbsp;Mengzheng Lv,&nbsp;Dongxue Zhang","doi":"10.1002/for.70101","DOIUrl":"https://doi.org/10.1002/for.70101","url":null,"abstract":"<div>\u0000 \u0000 <p>The accurate forecasting of Value at Risk (VaR) and Expected Shortfall (ES) is crucial for investment decisions, external regulation, and risk capital allocation. Existing literature has demonstrated that both high-frequency transaction data and economic policy uncertainty are important factors influencing volatility forecasts. Since tail risk measures are closely related to volatility, we propose fully extracting the information from 5-min transaction data and economic policy uncertainty to model tail risk dynamics. We extract various factors from 5-min transaction data that simulate volatility clustering, long memory, multi-scale behavior, jumps, overnight information, and leverage effects, all of which drive risk dynamics. Additionally, we explore the role of monthly economic policy uncertainty factors, which reflect various aspects of the macroeconomy in tail risk estimation. However, structural breaks caused by business cycles, extreme events, and changes in economic policies may lead to changes in the explanatory power of the driving factors over time. To address the instability of the model, this paper proposes a mixed-frequency time-varying combination framework that assigns time-varying weights to risk drivers of different frequencies to achieve forecast combination. The results from calibration tests, joint scoring functions, and Murphy diagrams consistently show that the proposed mixed-frequency time-varying combination framework outperforms existing models in terms of the accuracy and robustness of risk estimation.</p>\u0000 </div>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1681-1713"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148147234","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Meta-Learning Credit Risk Prediction by Fusing Regularized Logistic Regression and Random Forest","authors":"Min Zhang,&nbsp;Yong Li,&nbsp;Yiping Yang","doi":"10.1002/for.70097","DOIUrl":"https://doi.org/10.1002/for.70097","url":null,"abstract":"<div>\u0000 \u0000 <p>Credit risk assessment is of vital importance as it can predict whether an individual will default on a loan. Classifying loan applicants into “good” or “bad” debtors can help lending institutions make informed decisions. One challenge in credit default prediction is how to avoid the interference of redundant and irrelevant features, which may disrupt the classification results. This paper uses a meta-learning approach for credit risk prediction based on \u0000<span></span><math>\u0000 <msub>\u0000 <mrow>\u0000 <mi>L</mi>\u0000 </mrow>\u0000 <mrow>\u0000 <mn>1</mn>\u0000 </mrow>\u0000 </msub></math>/\u0000<span></span><math>\u0000 <msub>\u0000 <mrow>\u0000 <mi>L</mi>\u0000 </mrow>\u0000 <mrow>\u0000 <mn>2</mn>\u0000 </mrow>\u0000 </msub></math>-regularized logistic regression (RLR) and random forest (RF) to address this issue. The model reduces overfitting through RLR's weight constraint mechanism and models nonlinear feature dependencies using RF's ensemble of decision trees. A dynamic weight allocation mechanism in the meta-learning layer adaptively adjusts the predictive contributions of RLR and RF based on dataset characteristics, simultaneously optimizing feature weighting and model fusion strategies. We conducted an analysis of the German Credit dataset and the Lending Club dataset from three aspects: classification results, methodological differences, and interpretability. The experimental results demonstrate that elastic net–RF meta-learning (Enet–RF Meta) outperforms single models in accuracy, AUC, and F1 score. Feature importance analysis identifies key variables influencing predictions, while violin plots reveal varying reliance on linear features across datasets. The approach offers enhanced predictive performance and interpretability, making it a robust solution for credit risk assessment.</p>\u0000 </div>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1421-1437"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148146970","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adaptive Model Integration for Stock Price Forecasting With CNN-ATTN-BiLSTM","authors":"Yi Xiao,&nbsp;Chen He,&nbsp;Ming Yi,&nbsp;Yi Hu","doi":"10.1002/for.70098","DOIUrl":"https://doi.org/10.1002/for.70098","url":null,"abstract":"<div>\u0000 \u0000 <p>Stock price prediction is challenging due to the high volatility and complex nonlinear patterns in financial markets. Traditional time series forecasting methods often struggle to capture such intricate dynamics. To address this, we propose a novel adaptive model fusion framework that integrates Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory networks (BiLSTM), and Attention Mechanisms (ATTN) to improve stock price forecasting. Our approach employs multi-scale feature extraction using multiple CNNs enhanced with attention to adaptively select relevant features across different time horizons. An adaptive fusion mechanism dynamically adjusts the contribution of each sub-model according to input data, optimizing predictions under varying market conditions. We investigated the effects of both univariate and multivariate data on model performance, and analyze how data distribution characteristics influence forecasting accuracy. Experiments were conducted on stock data from the top 9 companies in the Nasdaq 100 index by market capitalization, validating the robustness and effectiveness of our method across different sectors. Results show that our model significantly outperformed traditional forecasting approaches, achieving higher accuracy and improved generalization in diverse market environments. This study offers a novel framework for stock price prediction and provides valuable insights into adaptive model integration for financial time series forecasting.</p>\u0000 </div>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1331-1349"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148147071","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Overnight Trading Matters!—Volatility Forecast in the Crude Oil Futures Market","authors":"Jing Hao,&nbsp;Feng He,&nbsp;Liyuan Qin","doi":"10.1002/for.70089","DOIUrl":"https://doi.org/10.1002/for.70089","url":null,"abstract":"<div>\u0000 \u0000 <p>This study primarily explores the potential of night-time trading information to enhance the prediction accuracy of WTI crude oil futures price volatility within the HAR framework. Out-of-sample analysis reveals that integrating daily, monthly, or weekly overnight information into the HAR models notably improves their capacity to forecast crude oil futures price volatility. This suggests that overnight information effectively improves forecasting performance. Moreover, we demonstrate that short-term overnight information possesses predictive power surpassing that of long-term information. Finally, these findings are robust to the application of the recursive window method and the HAR-CJ model.</p>\u0000 </div>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1511-1528"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148147253","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Universal Multivariate Long-Term Time-Series Robust Forecasting Model With Distinguishable Variable Identifier","authors":"Xin-ji Chen,&nbsp;Yang-yang He,&nbsp;Jian-wei Liu,&nbsp;Ze-yu Liu,&nbsp;Chao-dong Tan","doi":"10.1002/for.70105","DOIUrl":"https://doi.org/10.1002/for.70105","url":null,"abstract":"<div>\u0000 \u0000 <p>Recent studies have shown that, in multivariate long-term time-series forecasting, linear models employing a channel-independent (CI) strategy tend to outperform Transformer-based models that do not explicitly model cross-channel interactions, including FEDformer, Autoformer, and Informer. This finding has cast doubt on the ability of the Transformer's attention mechanism to capture temporal dependencies effectively. To address this issue, the Transformer-based PatchTST model also adopts the CI strategy. However, follow-up work has revealed that CI models suffer from the drawback of “spatial indistinguishability”; that is, they collapse several variables into an identical forecast whenever the past observations of those variables look alike, even if the variables later evolve in opposite directions. To overcome this limitation, we attach learnable variable identifiers to the embeddings of multivariate time series, enabling the model to differentiate individual variables. We further observe that these identifiers can be leveraged to analyze inter-variable similarities. Additionally, we employ multi-scale CNNs in the shallow layers to extract rich local temporal features while reducing computational overhead, and a Transformer sub-network to capture long-term dependencies. By combining the strengths of both components, the model extracts both local and global structures inherent in multivariate time series. Moreover, to enhance robustness against noise and outliers and to mitigate overfitting, we introduce a novel loss function that integrates the advantages of mean squared error (MSE) and mean absolute error (MAE). Extensive experiments on six widely used open-source datasets demonstrate that our model consistently outperforms the Transformer-based baseline (PatchTST), achieving a maximum relative improvement of 4.4%.</p>\u0000 </div>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1617-1632"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148146973","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Machine Learning Approaches to Forecast the Realized Volatility of Crude Oil Prices","authors":"Talha Omer,&nbsp;Kristofer Månsson,&nbsp;Pär Sjölander,&nbsp;Gazi Salah Uddin","doi":"10.1002/for.70107","DOIUrl":"https://doi.org/10.1002/for.70107","url":null,"abstract":"<p>This paper presents an evaluation of the accuracy of machine learning (ML) techniques in forecasting the realized volatility of West Texas Intermediate (WTI) crude oil prices. We compare several ML algorithms, including regularization, regression trees, random forests, and neural networks, to several heterogeneous autoregressive (HAR) models. The results show that the ML and regularization methods are efficient, even when there are only three predictors: daily, weekly, and monthly realized volatility (RV) lags. In addition, when the ML and regularization methods are applied, the results become more pronounced over longer forecasting horizons, as well as for weekly and monthly horizons. These ML methods are effective in approximating long-term realized volatility. Furthermore, we found that additional explanatory variables for real-time currency exchange contain valuable information to forecast the RV of crude oil prices.</p>","PeriodicalId":47835,"journal":{"name":"Journal of Forecasting","volume":"45 4","pages":"1633-1651"},"PeriodicalIF":2.7,"publicationDate":"2026-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/for.70107","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148147095","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}