Forecast Accuracy Research Articles

Enhancing Hierarchical Sales Forecasting with Promotional Data: A Comparative Study Using ARIMA and Deep Neural Networks

Retailers depend on accurate sales forecasts to effectively plan operations and manage supply chains. These forecasts are needed across various levels of aggregation, making hierarchical forecasting methods essential for the retail industry. As competition intensifies, the use of promotions has become a widespread strategy, significantly impacting consumer purchasing behavior. This study seeks to improve forecast accuracy by incorporating promotional data into hierarchical forecasting models. Using a sales dataset from a major Portuguese retailer, base forecasts are generated for different hierarchical levels using ARIMA models and Multi-Layer Perceptron (MLP) neural networks. Reconciliation methods including bottom-up, top-down, and optimal reconciliation with OLS and WLS (struct) estimators are employed. The results show that MLPs outperform ARIMA models for forecast horizons longer than one day. While the addition of regressors enhances ARIMA’s accuracy, it does not yield similar improvements for MLP. MLPs present a compelling balance of simplicity and efficiency, outperforming ARIMA in flexibility while offering faster training times and lower computational demands compared to more complex deep learning models, making them highly suitable for practical retail forecasting applications.

Exploring the integration of a global AI model with traditional data assimilation in weather forecasting

Abstract Recent advancements in artificial intelligence (AI) have profoundly transformed weather forecasting, challenging traditional reliance on numerical weather prediction (NWP) models. Despite notable progress, AI models still depend heavily on traditional NWP systems and data assimilation methods to generate analysis fields, a dependency that increases computational demands and might limit forecast accuracy. This study explored the integration of Gridpoint Statistical Interpolation (GSI) with the Pangu-Weather AI forecasting model (GSI-Pangu), and assessed the potential for AI models to autonomously generate forecasts by leveraging mature data assimilation systems. Our experiments commenced by adopting ERA5 reanalysis data for the initial cycle, and then involved assimilation of simulated observations in subsequent cycles, spanning a month-long period. Results demonstrated notable enhancements in forecast accuracy, with reductions in the root mean square error across various atmospheric variables compared with the results of a control experiment without data assimilation. Additionally, the results highlighted GSI-Pangu’s ability to predict large-scale circulation patterns of extreme precipitation events, together with its effectiveness in driving regional models to accurately forecast precipitation intensity and distribution. Successful implementation of GSI within the Pangu-Weather framework underscores the transformative potential of hybrid forecasting systems, which merge conventional meteorological techniques with AI innovations, thereby facilitating accelerated adoption of AI in weather forecasting.

Forecasting Gate-Front Water Levels Using a Coupled GRU–TCN–Transformer Model and Permutation Entropy Algorithm

Water level forecasting has significant impacts on transportation, agriculture, and flood control measures. Accurate water level values can enhance the safety and efficiency of water conservancy hub operation scheduling, reduce flood risks, and are essential for ensuring sustainable regional development. Addressing the nonlinearity and non-stationarity characteristics of gate-front water level sequences, this paper introduces a gate-front water level forecasting method based on a GRU–TCN–Transformer coupled model and permutation entropy (PE) algorithm. Firstly, an analysis method combining Singular Spectrum Analysis (SSA) and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is used to separate the original water level data into different frequency modal components. The PE algorithm subsequently divides each modal component into sequences of high and low frequencies. The GRU model is applied to predict the high-frequency sequence part, while the TCN–Transformer combination model is used for the low-frequency sequence part. The forecasting from both models are combined to obtain the final water level forecasting value. Multiple evaluation metrics are used to assess the forecasting performance. The findings indicate that the combined GRU–TCN–Transformer model achieves a Mean Absolute Error (MAE) of 0.0154, a Root Mean Square Error (RMSE) of 0.0205, and a Coefficient of Determination (R2) of 0.8076. These metrics indicate that the model outperforms machine learning Support Vector Machine (SVM) models, GRU models, Transformer models, and TCN–Transformer combination models in forecasting performance. The forecasting results have high credibility. This model provides a new reference for improving the accuracy of gate-front water level forecasting and offers significant insights for water resource management and flood prevention, demonstrating promising application prospects.

Applied Research on the Impact of Big Data, Cloud Computing and Cloud Storage on Smart Manufacturing Supply Chain

The rapid advancements in big data, cloud computing, and cloud storage technologies are profoundly reshaping the operational models of smart manufacturing supply chains. The integration and application of these technologies not only drive the enhanced data collection and analysis capabilities across various supply chain segments but also significantly improve the accuracy of demand forecasting, the rationality of resource allocation, and the scientific nature of decision support. Consequently, the collaborative efficiency and dynamic response capabilities of supply chains have been substantially elevated, providing robust technological backing for enterprises in navigating the intricately volatile market environment. Nonetheless, while reaping the benefits of these technological advancements, challenges such as data security and privacy protection, system integration, and technical selection cannot be overlooked. By judiciously selecting technology platforms, reinforcing data security management, and continuously investing in technological advancements and talent cultivation, the holistic effectiveness of smart manufacturing supply chains will be further optimized.

Using Machine Learning for Climate Modelling: Application of Neural Networks to a Slow-Fast Chaotic Dynamical System as a Case Study

This paper explores the capabilities of two types of recurrent neural networks, unidirectional and bidirectional long short-term memory networks, to build a surrogate model for a coupled fast–slow dynamic system and predicting its nonlinear chaotic behaviour. The dynamical system in question, comprising two versions of the classical Lorenz model with a small time-scale separation factor, is treated as an atmosphere–ocean research simulator. In numerical experiments, the number of hidden layers and the number of nodes in each hidden layer varied from 1 to 5 and from 16 to 256, respectively. The basic configuration of the surrogate model, determined experimentally, has three hidden layers, each comprising between 16 and 128 nodes. The findings revealed the advantages of bidirectional neural networks over unidirectional ones in terms of forecasting accuracy. As the forecast horizon increases, the accuracy of forecasts deteriorates, which was quite expected, primarily due to the chaotic behaviour of the fast subsystem. All other things being equal, increasing the number of neurons in hidden layers facilitates the improvement of forecast accuracy. The obtained results indicate that the quality of short-term forecasts with a lead time of up to 0.75 model time units (MTU) improves most significantly. The predictability limit of the fast subsystem (“atmosphere”) is somewhat greater than the Lyapunov time.

Deep-learning-based multi-timescale load forecasting in buildings: opportunities and challenges from research to deployment

Electricity load forecasting for buildings and campuses is becoming increasingly important as the penetration of distributed energy resources (DERs) grows. Efficient operation and dispatch of DERs require reasonably accurate predictions of future energy consumption in order to conduct near-real-time optimized dispatch of on-site generation and storage assets. Electric utilities have traditionally performed load forecasting for load pockets spanning large geographic areas, and therefore, forecasting has not been a common practice by buildings and campus operators. Given the growing trends of research and prototyping in the grid-interactive efficient buildings domain, characteristics beyond simple algorithm forecast accuracy are important in determining the algorithm’s true utility for smart buildings. Other characteristics include the overall design of the deployed architecture and the operational efficiency of the forecasting system. In this work, we present a deep-learning-based load forecasting system that predicts the building load at 1-hour intervals for 18 hours in the future. We also discuss challenges associated with the real-time deployment of such systems as well as the research opportunities presented by a fully functional forecasting system that has been developed within the National Renewable Energy Laboratory’s Intelligent Campus program.

Tropical Interbasin Interaction as Effective Predictors of Late-Spring Precipitation Variability in the Southern Great Plains

Abstract The southern Great Plains experience fluctuating precipitation extremes that significantly impact agriculture and water management. Despite ongoing efforts to enhance forecast accuracy, the underlying causes of these climatic phenomena remain inadequately understood. This study elucidates the relative influence of the tropical Pacific and Atlantic basins on April–May–June precipitation variability in this region. Our partial ocean assimilation experiments using the Community Earth System Model unveil the prominent role of interbasin interaction, with the Pacific and Atlantic contributing approximately 70% and 30%, respectively, to these interbasin contrasts. Our statistical analyses suggest that these tropical interbasin contrasts could serve as a more reliable indicator for late-spring precipitation anomalies than El Niño–Southern Oscillation. The conclusions are reinforced by analyses of seven climate forecasting systems within the North American Multi-Model Ensemble, offering an optimistic outlook for enhancing real-time forecasting of late-spring precipitation in the southern plains. However, the current predictive skills of the interbasin contrasts across the prediction systems are hindered by the lower predictability of the tropical Atlantic Ocean, pointing to the need for future research to refine climate prediction models further. Significance Statement Agriculture and infrastructure in the southern plains face challenges from severe late-spring precipitation extremes. Traditional predictors like El Niño–Southern Oscillation (ENSO) lose effectiveness during the critical spring-to-summer transition, creating a forecasting gap. This study introduces the concept of tropical interbasin interactions, known to enhance seasonal predictability for late-spring precipitation in the southern plains. Novel climate model experiments highlight contributions from the tropical Pacific and Atlantic, offering a promising predictability that potentially surpasses the limitations of ENSO-based predictions. These outcomes hold the potential for developing operational forecasts of late-spring precipitation anomalies in the southern plains, enabling proactive risk management.

FuXi‐En4DVar: An Assimilation System Based on Machine Learning Weather Forecasting Model Ensuring Physical Constraints

AbstractRecent machine learning (ML)‐based weather forecasting models have improved the accuracy and efficiency of forecasts while minimizing computational resources, yet still depend on traditional data assimilation (DA) systems to generate analysis fields. Four dimensional variational data assimilation (4DVar) enhances model states, relying on the prediction model to propagate observation to the initial field. Consequently, the initial fields from traditional DA are not optimal for ML‐based models, necessitating a customized DA system. This paper introduces an ensemble 4DVar system integrated with the FuXi model (FuXi‐En4DVar), which can independently generate accurate analysis fields. It utilizes automatic differentiation to compute gradients, and demonstrates the equivalence of these gradients with those derived from adjoint models. Experimental results indicate that this system preserves the physical balance of the analysis field and exhibits flow‐dependent characteristics. These features enhance the propagation and assimilation of observation into the initial analysis field, thereby improving the accuracy of the analysis fields.

Robust Estimation for Threshold Autoregressive Moving-Average Models

Threshold autoregressive moving-average (TARMA) models extend the popular TAR model and are among the few parametric time series specifications to include a moving average in a nonlinear setting. The state dependent reactions to shocks is particularly appealing in Economics and Finance. However, no theory is currently available when the data present heavy tails or anomalous observations. Here we provide the first theoretical framework for robust M-estimation for TARMA models and study its practical relevance. Under mild conditions, we show that the robust estimator for the threshold parameter is super-consistent, while the estimators for autoregressive and moving-average parameters are strongly consistent and asymptotically normal. The Monte Carlo study shows that the M-estimator is superior, in terms of both bias and variance, to the least squares estimator, which can be heavily affected by outliers. The findings suggest that robust M-estimation should be generally preferred to the least squares method. We apply our methodology to a set of commodity price time series; the robust TARMA fit presents smaller standard errors and superior forecasting accuracy. The results support the hypothesis of a two-regime non-linearity characterized by slow expansions and fast contractions.

Application of the Standardized Precipitation Index for Forecasting Seasonal Drought in Vietnam

Drought is a natural disaster that causes significant damages to the economy, society and environment. Forecasting droughts is crucial for disaster risk management and water security. This study introduces a new method for forecasting seasonal droughts in Vietnam using the Standardized Precipitation Index (SPI) which is based on a combination of dynamic modeling and statistical methods. The precipitation forecast products from regional climate models (RegCM, clWRF) with inputs from Climate Forecast System (CFSv2) were used to calculate the SPI index after statistical calibration. Results show that calibrating model precipitation with statistical methods such as Artificial Neural Networks (ANN) and Multivariate Linear Regression (MLR) significantly improves the precipitation forecast accuracy. Evaluation indices indicate that the SPI calculated with the post-calibration model precipitation has a good reproducibility for drought events, particularly mild droughts. However, there remains some limitations on accurate forecast of the intensity and spatial distribution of moderate drought events in certain regions such as Central Vietnam.

Applying machine learning to electricity price forecasting in simulated energy market scenarios

Policy packages, such as the “European Green Deal”, call for a substantial restructuring of the power plant park. This, in combination with more flexible demand, will result in novel electricity price dynamics. These can be studied using, e.g., agent-based models which simulate bidding decisions of market actors, thereby uncovering emergent market phenomena. For their bidding decisions, simulated actors – just like real-world actors – require accurate market price forecasts. Techniques to obtain such forecasts need to be applicable to vastly different future electricity market scenarios, ideally without the need of scenario-specific retraining. This is a major difference compared to real-world electricity market forecasting, which is based on minimal variations in the underlying energy system. Despite the long track record in this field, it is not sufficiently clear which methods are suitable for forecasting simulated future electricity markets in greatly varying scenarios and technology mixes. To address this gap, we assess the applicability of different forecasting techniques to price time series generated by simulations of the future electricity market. We then evaluate the forecast accuracy of two recent machine learning architectures, namely N-BEATS and Temporal Fusion Transformers, based on parameter combinations with significant expansions of renewable energy and flexibility option capacity. As expected, the results demonstrate that machine learning exhibits superior accuracy compared to naïve benchmarks. Particularly, when future covariates, such as residual load, are employed, the mean absolute error consistently remains below 1.40 EUR/MWh. This may be attributed to reduced inner complexity of simulated electricity prices compared to real-world market dynamics. Our findings demonstrate that machine learning can provide reliable forecasts of future electricity prices and that retraining may not be necessary even with widely varying shares of renewable energy and flexibility capacity. These forecasting methods could therefore be effectively employed in electricity market simulations in the context of the energy transition.

Electric Vehicle Energy Management via Traffic Light Detection and Segmental Velocity Forecasting

Abstract Predictive-based power control has been widely recognized as a promising approach to boost driving range and improve system-level energy efficiency for electric vehicles (EVs), in which vehicle velocity forecasting generally serves as a preliminary input to optimally schedule the operations of varying onboard electrical and thermal systems. A segment-based velocity forecasting approach for individual commuting vehicles developed in this study reveals that it is challenging to forecast the velocity at intersection segments only using the velocity data. To address this challenge, this study seeks to develop a YOLO-V2-based object detection deep network to recognize the traffic lights in advance and leverage the detected signals to establish a forecasting model that integrates with the probability-based hybrid forecasting approach. The case study results show that the traffic light detection-based forecasting model can significantly improve the forecasting accuracy for intersection segments. Based on the forecasting velocity 5–15 s ahead, the effectiveness of model predictive control-based energy management strategy is further evaluated with a liquid-based battery thermal control system. The proposed battery thermal management system (BTMS) model shows promising results in maintaining battery temperature within an appropriate range, thus improving the overall energy efficiency of the EV. Moreover, a traffic light-based real-time energy management framework is developed to directly control the power demand from the air conditioning (AC) system.

Nowcasting GDP using machine learning methods

Abstract This paper compares the ability of several econometric and machine learning methods to nowcast GDP in (pseudo) real-time. The analysis takes the example of Dutch GDP over the period 1992Q1–2018Q4 using a broad data set of monthly indicators. It discusses the forecast accuracy but also analyzes the use of information from the large data set of macroeconomic and financial predictors. We find that, on average, the random forest provides the most accurate forecast and nowcasts, whilst the dynamic factor model provides the most accurate backcasts.

Simulation of time-invariant three-dimensional Freak waves and their scattering identification based on the JONSWAP spectrum and Donelan direction function

Abstract The results of recent maritime accident investigations show that many maritime accidents are related to the attack of freak waves, which can sink ships instantaneously with colossal energy. To identify and forecast freak waves accurately and avoid the danger of freak waves to marine structures and people, we have conducted further research on freak waves. In this paper, we propose an efficient time-invariant three-dimensional (3-D) freak wave computational model to study the scattering characteristics of time-invarian 3-D freak waves and obtain a more explicit identification of their scattering differences. First, this paper adopts linear superposition method to simulate a 3-D random rough sea surface based on the JONSWAP wave spectrum and Donelan directional distribution function. Moreover, we analyze the effect of different angular frequencies on the simulated sea surface. At the same time, the Monte Carlo method is used to further simulate the capillary waves based on the random sea surface, and this method vastly improves the computational accuracy of the random sea surface. Then, based on the two-wave train superposition wave energy focusing mode, a numerical calculation method of time-invariant 3-D freak waves is proposed, and the time-series evolution process of 3-D freak waves is simulated. Finally, the backscattering coefficients of the 3-D freak wave surface are calculated using the two-scale method, and the electromagnetic scattering characteristics of the freak wave surface under different evolution stages, wind speeds, and radar incidence angles are analyzed. By this method, we conclude that the recognition of freak waves is more accurate under medium-high wind speed and significant incidence angle conditions. The experiments show that the research in this paper can identify and predict freak waves more effectively, further improve the forecast accuracy of future freak waves, and also have some application value for freak wave risk warnings.

Integrated Reporting, ESG Disclosure, Forecast Accuracy, and Firm Value: Profitability as Moderating Variable

Objective: This study investigate how the quality of IR and ESG disclosures can affect the level of information asymmetry and enhance legitimacy and signal about firm value. Theoretical Framework: IR will increase the company's transparency to stakeholders so that reduces the problem of information asymmetry and long-term business value creation. SDGs encourages organizations to pursue credentials by implementing ESG to increase corporate credibility. High profitability can provide positive signals that encourage managers to implement IR and submit ESG reports, thereby reducing information asymmetry and increasing firm value. Method: Based on data from the IDX database, this study analyses companies that are members of the LQ45 group. We have collected annual reports and sustainability reports for the period between 2013 and 2022. Results and Discussion: IR quality has a negative influence on FA and firm value, while ESG disclosure has a positive influence on FA and firm value. Profitability variable is able to strengthen the effect of IR quality and ESG disclosure on FA. Therefore, IR quality and ESG disclosures have a strong influence on FA. Research Implications: Theoretically, this study’s result strengthen the study of legitimacy and institutional theories in explaining the relationship between IR quality and sustainability disclosures on information asymmetry and firm value in the long run. Practically, it highlights the importance for managers to enhance IR and ESG disclosures, thereby promoting corporate transparency and accountability. Originality/Value: This considers profitability variables to understand the motives and drivers of managers to implement integrated reporting formats and deliver sustainability reporting.

A Study on a Novel Production Forecasting Method of Unconventional Oil and Gas Wells Based on Adaptive Fusion

Reliable forecasting of unconventional oil and gas well production has consistently been a hot and challenging issue. Most existing data-driven production forecasting models rely solely on a single methodology, with the application effects of other mainstream algorithms remaining unclear, which to some extent hinders the generalization and utilization of these models. To address this, this study commences with data preparation and systematically develops a novel forecasting model based on the adaptive fusion of multiple mainstream data-driven algorithms such as random forest and support vector machine. The validity of the model is verified using actual production wells in the Marcellus. A comprehensive evaluation of multiple feature engineering extraction techniques concludes that the main controlling factors affecting the production of Marcellus gas wells are horizontal segment length, fracturing fluid volume, vertical depth, fracturing section, and reservoir thickness. Evaluation models based on these primary controlling factors reveal significant differences in prediction performance among mainstream data-driven methods when applied to the dataset. The newly developed model based on adaptive fusion optimized by genetic algorithms outperforms individual models across various evaluation metrics, which can effectively improve the accuracy of production forecasting, demonstrating its potential for promoting the application of data-driven methods in forecasting unconventional oil and gas well production. Furthermore, this will assist enterprises in allocating resources more effectively, optimizing extraction strategies, and reducing potential costs stemming from inaccurate predictions.

Enhancing the Predictability of Wintertime Energy Demand in The Netherlands Using Ensemble Model Prophet-LSTM

Energy demand forecasting is crucial for maintaining stable and affordable energy supplies, especially for vulnerable populations most affected by shortages and high costs. In the Netherlands, transmission system operator TenneT has raised concerns about potential electricity shortages by 2030. Rising energy prices and the impact of climate change on the energy demand further complicate today’s energy market. Policymakers lack clear insights into demand patterns, which complicates the optimization of energy use and the protection of at-risk communities. Accurate and timely forecasts are essential for addressing these issues and supporting sustainable energy management. This research focuses on enhancing the accuracy and lead time of wintertime energy demand forecasts in the Netherlands using advanced machine learning. The ensemble model Prophet-LSTM is trained on hourly load consumption data combined with climate change-related and energy price predictors. The results demonstrate significant improvements over baseline models, achieving a Pearson correlation coefficient of r=0.93 compared to r=0.50 in prior studies, as well as accurate forecasts up to 180 days ahead, compared to 2 months. Incorporating climate change-related predictors is challenging due to multicollinearity, highlighting the importance of careful predictor selection. Including energy price predictors yielded modest yet hopeful results, suggesting their ability to optimize energy demand forecasting.

Validity of Laws for Predicting Lifespan of Welded Points Under Variable Loading

The article examines the validity of the laws for predicting the life of welded points under variable amplitude loading, using three steels (HE360D, XE360D and XES) supplied by ArcelorMittal to Renault for the manufacture of automobile chassis parts. The fatigue tests revealed two modes of failure: cracking of the sheets at medium stresses and shearing of the molten core at high forces. For lifespan prediction, three damage laws were studied: Miner's law, Mesmacque and Amrouche law, and a proposed model. Miner's law, despite its simplicity and wide distribution, shows notable limitations in the presence of variable loads, because it does not take into account the interactions between successive load levels. The law of Mesmacque and Amrouche uses a damage indicator linked to the FN curve of the material, allowing a cycle by cycle estimation of the residual lifespan. However, it remains non-conservative at low effort. The proposed model, based on Chaboche law, overcomes some of its theoretical drawbacks, notably the dependence on the calibration parameters of the SN curve. By fitting this curve within a specific calibration window, the model improves forecast accuracy by taking into account parameter variability. This proposed model provides more reliable lifespan predictions that better correlate with experimental results, especially for low effort levels where Miner's law fails. The validation approach is based on comparing the predictions of the damage laws with the experimental results, by evaluating the deviation from the first bisector of the graphs. The results show that Miner's law is unsuitable for effort levels, while the proposed model turns out to be the most precise and conservative. The law of Mesmacque and Amrouche falls between the two, presenting a certain justice but remaining overall non-conservative. In conclusion, the proposed model offers a better match with experimental data, making its predictions more reliable for industrial applications.

The Impact of ESG Reporting on Analyst Forecast Accuracy: A Cross-Market Review

The Impact of ESG Reporting on Analyst Forecast Accuracy: A Cross-Market Review

Specification Choices in Quantile Regression for Empirical Macroeconomics

ABSTRACTQuantile regression has become widely used in empirical macroeconomics, in particular for estimating and forecasting tail risks. This paper examines various choices in the specification of quantile regressions for macro applications, including how and to what extent to include shrinkage and whether to apply shrinkage in a classical or Bayesian framework. We focus on forecasting accuracy, measured with quantile scores and quantile‐weighted continuous ranked probability scores at a range of quantiles from the left to right tail. Across applications, we find that shrinkage is generally helpful to quantile forecast accuracy, with Bayesian quantile regression dominating frequentist quantile regression.JEL Classification: C53, E17, E37, F47