Journal of Forecasting最新文献

We examine the role of the technology in predicting carbon prices using a large set of machine learning models. The predictors are selected from technological, environmental, financial, energy, and geopolitical aspects. Our sample covers the daily period from August 1, 2014, to March 4, 2024. We find that technology factors (Information Technology Index, AEX Technology Index, and Tech All Share Index) significantly improve the prediction accuracy of carbon prices, both when included in the prediction model individually and simultaneously. Furthermore, the Diebold–Mariano and Clark–West tests highly reject the null of equal predictive accuracy between the technology model and the baseline model (without technology variables). Moreover, results show that XGBoost outperforms the alternative machine learning models for all forecasting horizons (1, 5, 22, and 250 days). We present significant policy implications useful for investors, companies, and policymakers.

This article proposed a novel hybrid framework, the WTC-DCA-Informer, for forecasting volatility in the Australian stock market. The findings indicated that (1) through a comprehensive comparison with various machine learning and deep learning models, the proposed WTC-DCA-Informer framework significantly outperformed traditional methods in terms of predictive performance. (2) Across different training set proportions, the WTC-DCA-Informer model demonstrated exceptional forecasting capabilities, achieving a coefficient of determination (R²) as high as 0.9216 and a mean absolute percentage error (MAPE) as low as 13.6947%. (3) The model exhibited strong adaptability and robustness in responding to significant market fluctuations and structural changes before and after the outbreak of COVID-19. This study offers a new perspective and tool for forecasting financial market volatility, with substantial theoretical and practical implications for enhancing the efficiency and stability of financial markets.

This paper proposes a novel hybrid nonlinear interval decomposition ensemble (NIDE) framework to improve forecasting accuracy of interval-valued gas prices. The framework first decomposes the price series using bivariate empirical mode decomposition and interval multiscale permutation entropy to capture dynamics driven by long-term trends, events, and short-term fluctuations. Tailored models are then employed for each component, including a threshold autoregressive interval model, interval event study methodology, and interval random forest. Finally, an ensemble prediction integrates the component forecasts. Empirical results show that the NIDE approach significantly outperforms benchmarks in out-of-sample forecasting of interval-valued natural gas prices. For instance, the RMSE improvements range from 10.3% to 38.8% compared to benchmark models. Additionally, the NIDE approach not only enhances accuracy but also provides economic interpretation by identifying drivers like speculative trading and public interest proxied by online trends.

Interval-valued data are a special symbolic data, which contains rich information. The prediction of interval-valued data is a challenging task. In terms of predicting interval-valued data, machine learning algorithms typically consider mean regression, which is sensitive to outliers and may lead to unreliable results. As an important complement to mean regression, in this paper, a quantile regression artificial neural network based on a center and radius method (QRANN-CR) is proposed to address this problem. Numerical studies have been conducted to evaluate the proposed method, comparing with several traditional models, including the interval-valued quantile regression, the center method, the MinMax method, and the bivariate center and radius method. The simulation results demonstrate that the proposed QRANN-CR model is an effective tool for predicting interval-valued data with higher accuracy and is more robust than the other methods. A real data analysis is provided to illustrate the application of QRANN-CR.

Real-world systems are often formulated as constrained optimization problems. Techniques to incorporate constraints into neural networks (NN), such as neural ordinary differential equations (Neural ODEs), have been used. However, these introduce hyperparameters that require manual tuning through trial and error, raising doubts about the successful incorporation of constraints into the generated model. This paper describes in detail the two-stage training method for Neural ODEs, a simple, effective, and penalty parameter-free approach to model constrained systems. In this approach, the constrained optimization problem is rewritten as two optimization subproblems that are solved in two stages. The first stage aims at finding feasible NN parameters by minimizing a measure of constraints violation. The second stage aims to find the optimal NN parameters by minimizing the loss function while keeping inside the feasible region. We experimentally demonstrate that our method produces models that satisfy the constraints and also improves their predictive performance, thus ensuring compliance with critical system properties and also contributing to reducing data quantity requirements. Furthermore, we show that the proposed method improves the convergence to an optimal solution and improves the explainability of Neural ODE models. Our proposed two-stage training method can be used with any NN architectures.

This study proposes a novel deep auto-optimized architecture for stock price forecasting that integrates sectoral behavior with individual stock sentiment to improve predictive accuracy. Traditional stock prediction models often focus solely on individual stock behavior, overlooking the impact of broader sectoral trends. The proposed approach utilizes advanced deep learning models, including gated recurrent units (GRU), bidirectional GRU, long short-term memory (LSTM), and bidirectional LSTM, with their hybrid ensembles. These models are built using the Keras functional API and auto ML network architecture search technology. The current deep auto-optimized multimodal framework incorporates sectoral behavior, significantly improving performance metrics. This research highlights the critical role of integrating sectoral behavior in stock price prediction models.

This paper introduces a novel spread option pricing model, the nonparametric predictive inference–based copula spread option model (NPIC-SOM), designed to evaluate the interdependence of multiple underlying assets. Through empirical analysis focused on Brent-WTI spread options, a widely traded derivative, we compare the predictive performance of the NPIC-SOM against the traditional geometric Brownian motion crack spread option model (GBM-CSOM). Our findings reveal that the NPIC-SOM not only forecasts spread option prices closer to empirical values but also captures market fluctuations more accurately than the GBM-CSOM. This superiority extends across various option types, moneyness levels and delta hedge efficiency. Furthermore, the NPIC-SOM's reliance on time-varying parameters enhances prediction accuracy, particularly for extreme market scenarios. These results indicate the practicality and efficiency of the NPIC-SOM as a robust spread option pricing model, offering valuable insights for option pricing strategies in financial markets.