Forecast Accuracy Research Articles

СОВРЕМЕННОЕ СОСТОЯНИЕ СИСТЕМЫ ГИДРОЛОГИЧЕСКИХ НАБЛЮДЕНИЙ

The article considers the key aspects of optimization of the composition and placement of the hydrological network of the Republic of Kazakhstan in the context of ensuring sustainable monitoring of water resources. The main attention is paid to the preservation of reference (long-row) observation points necessary for territorial generalizations, development of methods for hydrological calculations and forecasts, as well as assessment of long-term changes in the hydrological regime in the context of climate change. The scientific study emphasizes the importance of increasing the number of information posts to improve the reliability of hydrological forecasts and warnings of hazardous phenomena, equipping the hydrological network with modern equipment in strategically important areas, and restoring specialized observation points for experimental research. The implementation of the proposed measures will meet the current and future needs of the economy for hydrological information, as well as improve the accuracy and reliability of hydrological calculations and forecasts.

STFM: Accurate Spatio-Temporal Fusion Model for Weather Forecasting

Meteorological prediction is crucial for various sectors, including agriculture, navigation, daily life, disaster prevention, and scientific research. However, traditional numerical weather prediction (NWP) models are constrained by their high computational resource requirements, while the accuracy of deep learning models remains suboptimal. In response to these challenges, we propose a novel deep learning-based model, the Spatiotemporal Fusion Model (STFM), designed to enhance the accuracy of meteorological predictions. Our model leverages Fifth-Generation ECMWF Reanalysis (ERA5) data and introduces two key components: a spatiotemporal encoder module and a spatiotemporal fusion module. The spatiotemporal encoder integrates the strengths of convolutional neural networks (CNNs) and recurrent neural networks (RNNs), effectively capturing both spatial and temporal dependencies. Meanwhile, the spatiotemporal fusion module employs a dual attention mechanism, decomposing spatial attention into global static attention and channel dynamic attention. This approach ensures comprehensive extraction of spatial features from meteorological data. The combination of these modules significantly improves prediction performance. Experimental results demonstrate that STFM excels in extracting spatiotemporal features from reanalysis data, yielding predictions that closely align with observed values. In comparative studies, STFM outperformed other models, achieving a 7% improvement in ground and high-altitude temperature predictions, a 5% enhancement in the prediction of the u/v components of 10 m wind speed, and an increase in the accuracy of potential height and relative humidity predictions by 3% and 1%, respectively. This enhanced performance highlights STFM’s potential to advance the accuracy and reliability of meteorological forecasting.

Power data analysis and mining technology in smart grid

This study proposes a smart grid model named “GridOptiPredict”, which aims to achieve efficient analysis and processing of power system data through deep fusion of deep learning and graph neural network, so as to improve the intelligent level and overall efficiency of power grid operation. The model integrates three core functions of load forecasting, power grid state sensing and resource optimization into one, forming a closely connected and complementary framework. Through carefully designed experimental scheme, the practical value and effectiveness of “Grid OptiPredict” model are fully verified from three aspects: accuracy of load forecasting, sensitivity of power grid state sensing and efficiency of resource allocation strategy. Experimental results show that the model has significant advantages in prediction accuracy, model stability and robustness, resource optimization, security, information security, social and economic benefits and user experience.

Enhanced Foreign Exchange Volatility Forecasting Using Ceemdan with Optuna-Optimized Ensemble Deep Learning Model

Foreign Exchange (FX) is the largest financial market in the world, with a daily trading volume that significantly exceeds that of stock and futures markets. The prediction of FX volatility is a critical financial concern that has garnered significant attention from researchers and practitioners due to its far-reaching implications in the financial markets. This paper presents a novel hybrid ensemble forecasting model integrating a decomposition strategy and three deep learning (DL) models: Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), and Convolutional Neural Network (CNN). This combination addresses individual models' limitations and further improves the accuracy and stability of FX volatility forecasting. The proposed approach utilizes the CEEMDAN technique to decompose volatility into multiple distinct intrinsic mode functions (IMFs) and merges these IMFs with GARCH and EGARCH volatilities to form the input dataset for the DL models. In addition, we employed an attention mechanism to improve the effectiveness of the DL techniques. Furthermore, the hyperparameters for the DL models are optimized using the Optuna algorithm. Finally, a hybrid ensemble model for forecasting exchange rate volatility is developed by combining the predictions of three distinct DL models. The proposed approach is evaluated against various benchmark models using evaluation measures such as MSE, MAE, HMSE, HMAE, RMSE, Q-LIKE, and the model confidence set (MCS) approach. The results demonstrate that our proposed approach provides accurate and reliable forecasts of FX volatility under different forecasting regimes, making it a valuable tool for financial practitioners and researchers.

Comparative analysis of the predictive power of statistical models of macroeconomic indicators in conditions of permanent crises

Subject. The article discusses prediction of the state of the economy, the accuracy of forecasts of traditional models during crises, the need to find more effective model specifications to predict macro indicators. Objectives. The purpose is to carry out a comparative analysis of the predictive ability of ensemble methods in comparison with a set of models, including traditional statistical algorithms and machine learning algorithms. Methods. The comparative analysis of predictive ability of the models and interpretation of results obtained were performed using a dynamic factor model (DFM), a neural network with long-term short-term memory (LSTM), and integrated gradient methods (IG). Results. We performed the analysis of predictive ability of the ensemble model to forecast GDP, which combines DFM and LSTM to account for both linear and nonlinear dependencies in the data; the analysis of predictive power of various indicators, which showed that an increase in forecast error is observed for all models except DFM, the ensemble model with an error correction structure, and ARMAX. The obtained results can be used to build models of macroeconomic indicators in order to make strategic decisions by enterprises of various industries operating in a highly uncertain environment. Conclusions. The combination of DFM and LSTM in the ensemble provides higher accuracy forecasts than LSTM and competitor models, however, with less predictive power than DFM.

Wind power forecasting with metaheuristic-based feature selection and neural networks

Accurate forecasting of wind power generation is crucial for ensuring a stable and efficient energy supply, reducing the environmental impact of energy production, and promoting a cleaner and more sustainable energy supply. Inaccurate forecasts can lead to a mismatch between wind power generation and energy demand, resulting in wasted energy, increased emissions, and reduced grid stability. Therefore, improving the accuracy of wind power generation forecasting is essential for optimizing energy storage and grid management, reducing the reliance on fossil fuels, decreasing greenhouse gas emissions, and promoting a more sustainable energy future. This study proposes an innovative approach to enhance wind power generation forecasting accuracy by leveraging the strengths of metaheuristic algorithms for feature selection and integrating them with Neural Networks (NN). Specifically, five distinct algorithms - Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Teaching-Learning-Based Optimization (TLBO), and Evolutionary Mating Algorithm (EMA) - are integrated with NN model to identify optimal feature subsets from a comprehensive dataset of 18 diverse features. The results show that the GA consistently outperforms other algorithms in selecting the most influential features, leading to improved precision in wind power predictions. Notably, the GA achieves the best root mean square error (RMSE) of 37.1837 and the best mean absolute error (MAE) of 18.6313, outperforming the other algorithms and demonstrating the importance of feature selection in improving the accuracy of wind power forecasting. This innovative framework advances the field of renewable energy forecasting and provides valuable insights into optimizing feature sets for improved predictions across diverse domains.

Predicting directional volatility: HAR model with machine learning integration

ABSTRACT Empirical studies on volatility forecasting have predominantly concentrated on point or interval estimates. However, the study of directional changes in volatility remains a relatively underexplored domain. Prediction of directions in volatility is particularly important for timing strategies and asset allocation. In this paper, we consider the predictability of both short-term and long-term volatility directions using a range of machine learning techniques integrating a parsimonious Heterogeneous Autoregressive (HAR) structure. An empirical analysis was undertaken using the S&P500’s realized volatility and the Chicago Board Options Exchange’s (CBOE) VIX implied volatility. We show that the machine learning techniques consistently enhance the accuracy of volatility directional forecasts. Among these techniques, the Support Vector Machine (SVM) stands out by consistently achieving significant forecasting accuracy and the most substantial economic gains.

Novel Custom Loss Functions and Metrics for Reinforced Forecasting of High and Low Day-Ahead Electricity Prices Using Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) and Ensemble Learning

Day-ahead electricity price forecasting (DAEPF) is vital for participants in energy markets, particularly in regions with high integration of renewable energy sources (RESs), where price volatility poses significant challenges. The accurate forecasting of high and low electricity prices is particularly essential, as market participants seek to optimize their strategies by selling electricity when prices are high and purchasing when prices are low to maximize profits and minimize costs. In Japan, the increasing integration of RES has caused day-ahead electricity prices to frequently fall to almost zero JPY/kWh during periods of high RES output, creating significant profitability challenges for electricity retailers. This paper introduces novel custom loss functions and metrics specifically designed to improve the forecasting accuracy of extreme prices (high and low prices) in DAEPF, with a focus on the Japanese wholesale electricity market, addressing the unique challenges posed by the volatility of RES. To implement this, we integrate these custom loss functions into a Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) model, augmented by an ensemble learning approach and multimodal features. The proposed custom loss functions and metrics were rigorously validated, demonstrating their effectiveness in accurately predicting high and low electricity prices, thereby indicating their practical application in enhancing the economic strategies of market participants.

Integration of machine learning and deep learning methods for sunflower yield prediction

Practical experience in yield forecasting demonstrates that it is a complex multifactorial task requiring precise and reliable methods for resolution. The use of machine learning and deep learning is crucial for achieving better results in digital agronomy. This paper is dedicated to the development of an intelligent system for sunflower yield forecasting. Based on the analysis of scientific publications and practical experience, the main issues of data processing are summarized, and schemes for their resolution are proposed. The main stages of the work included the study of the current state of digital agronomy, the selection of an approach, the development of a data processing method, the software implementation of the model, and testing. The key challenges were the limited data sets and the complexity of choosing the optimal approach to avoid overfitting. Combining different analysis methods allowed the creation of a powerful system that surpasses traditional approaches, though it requires more data for training. The study's conclusions show that machine learning and deep learning methods, such as LightGBM and U-Net, along with the proposed data processing methods, achieve high accuracy in forecasting. The model demonstrated the ability to generalize knowledge to new fields and to build detailed yield maps. Further research includes the development of a method for generating combinations of plant care options, the adaptation of computer vision methods with optimized algorithms to reduce computational complexity, and the expansion of the system's functionality to include cybersecurity aspects. The proposed system significantly enhances the efficiency of sunflower yield forecasting, contributing to the development of digital agronomy.

Global atmospheric state analysis using objective Gaussian probability density functions

Atmospheric state analysis is a difficult scientific problem due to the chaotic nature of the atmosphere. Data assimilation is a framework for generating an accurate state analysis of a physical system using probability density functions (PDFs) describing uncertainty of information on the state of the physical system. However, since PDFs cannot be deduced theoretically, those used in data assimilation of atmospheric state analysis are based on empirical tunings. This PDF uncertainty limits the theoretical consistency and accuracy of atmospheric state analysis and that of all atmospheric sciences. In this study, we constructed a highly accurate and theoretically consistent atmospheric state analysis by objectively estimating the PDFs of all datasets (forecasts and observations) under the Gaussian approximation. We show that an ensemble of data assimilations with 192 members using the four-dimensional variational method and sample statistics obtained with the data assimilation theory (Desroziers’ method) can generate more accurate objective Gaussian PDFs, including flow-dependent forecast error structures. Numerical experiments of atmospheric state analysis and forecasts using objective PDFs were conducted and compared with those using conventional empirical PDFs. The objective PDFs had smaller error variances for most data (about 34% of those of CNTL on average) and larger observation error correlations for satellite radiances, where the strongest correlation was greater than 0.8. The analysed atmospheric states are systematically different, such as a cooler (exceeding 1.2 K) and wetter (exceeding 1.2 g/kg) low troposphere in regions characterized by low-level clouds off the west coast of the continents. The theoretical consistency evaluated by the chi-square-based tests showed a clear improvement from 16 to 95%. The forecast accuracy was improved globally up to 9%, with 95% statistical significance. The tropical cyclone track forecast accuracy was also improved about 20%.

Mineral policy and sustainable development goals: Volatility forecasting in the Global South's minerals market

This paper aims to evaluate the volatility of precious metals, specifically Palladium, Gold, and Platinum, within the context of the global minerals market. The research focuses on understanding the price dynamics of these metals and their implications for sustainable development, particularly in the Global South. The study employs a comprehensive approach, utilizing advanced machine learning and deep learning models such as GRU, Huber, Lasso, LSTM, Random Forest, Ridge Regression, SVM, ANN, and XGBoost. These models are assessed based on their forecasting accuracy for different time horizons, using metrics such as RMSE and MAPE. The findings reveal that the ANN, XGBoost, and LSTM models exhibit robust performance in forecasting the volatility of precious metals across various time horizons. The research highlights the unique volatility patterns of each metal and underscores the effectiveness of machine learning techniques in capturing these dynamics. The study acknowledges limitations such as the exclusion of macroeconomic and geopolitical factors in the forecasting models. Future research is suggested to integrate these factors to enhance forecasting accuracy. The study's findings are pivotal for investors, policymakers, and market regulators, especially in the context of the Global South and sustainable development. The research offers valuable insights for risk management strategies, investment planning, and policy formulation aimed at promoting market stability and sustainable economic growth. The study emphasizes the importance of selecting appropriate forecasting models based on specific time horizons and market requirements.

Development of a software service for stock price forecasting based on sentiment analysis and autoregressive models

This paper addresses the critical need for efficient market analysis tools in the era of big data and artificial intelligence. We present a novel software service that integrates real-time news sentiment analysis with stock market prediction, enhancing the accuracy and speed of trading decisions. The system employs APIs for data collection, FinBERT for sentiment analysis, and MongoDB for data storage, overcoming limitations of existing platforms like Investing.com and MarketWatch. Our methodology combines sentiment analysis with autoregressive models to forecast stock prices for 11 major companies. The experiment utilized 141 observations, applying multiple regression and binary outcome models. Results demonstrate that investor sentiment significantly affects stock prices for 2 out of 11 companies, with Meta showing a 70 % determination coefficient in price direction changes based on sentiment. The study reveals that incorporating both quantitative (previous stock prices) and qualitative (sentiment) data improves forecast accuracy for certain stocks. This research contributes to the field of financial analytics by providing a more comprehensive approach to stock price prediction, integrating ML models and data analytics to support informed decision-making in dynamic financial markets.

An adaptive method for real‐time photovoltaic power forecasting utilizing mathematics and statistics: Case studies in Australia and Vietnam

AbstractThe advancement of Photovoltaic technology has undergone rapid acceleration in recent years. Nonetheless, the most significant drawback of Photovoltaic is its intermittence, making it an obvious source of power fluctuation. This study proposes a novel scheme for real‐time or intraday PV power forecasting by adopting two predictive models, namely, White‐box and Combination. The White‐box model is implemented employing mathematical calculations and statistics called Exceedance Probability. Meanwhile, the Combination model is an aggregation of several predictive models' outputs including the White‐box model and benchmark ones by dynamically adjusting the weight coefficient of each model based on their forecasting accuracy. The experimental results, which are verified on two PV systems corresponding to two case studies located at Vietnam and Australia, indicate that the two proposed models outperform other referenced models as improves approximately 40% and 38% in terms of the first and second case study, respectively. In particular, the White‐box model shows superiority by updating the forecast every 10 min, which can adapt to the fluctuation of weather conditions whereas the Combination one yields acceptable precision, indicating its flexible application.

Enhancing climate forecasting with AI: Current state and future prospect

Background The escalating impact of climate change underscores the critical need for advanced and sustainable climate forecasting techniques. This review examines the current state and future prospects of leveraging Artificial Intelligence (AI) for climate forecasting, focusing on enhancing accuracy and identifying complex patterns in large datasets. Methods A systematic bibliometric methodology was employed, analyzing peer-reviewed literature from the past two decades. The study screened 455 articles from Scopus and Web of Science databases using specific keywords related to AI and weather forecasting. After removing duplicates and irrelevant studies, 218 articles were selected for detailed analysis. Bibliometric analysis was conducted using RStudio software to examine publication trends, co-word co-occurrence, and thematic evolution. Results The findings indicate significant growth in AI applications for climate forecasting, particularly from 2014 to 2023. AI techniques such as machine learning, artificial neural networks, and deep learning have shown promise in improving the accuracy of weather forecasts and early warning systems. The thematic analysis identified key themes like numerical weather prediction, feature selection, and neural networks as fundamental areas of research. Additionally, AI-based early warning systems for extreme weather events were highlighted as a crucial application. Below Figure 1. shows the graphical abstract of research Conclusions AI has the potential to significantly enhance climate forecasting by analyzing vast amounts of data and identifying complex patterns. Future research should focus on developing universal AI models, increasing model accuracy with explainable AI techniques, and integrating region-specific forecasts to aid decision-making in various sectors. Addressing ethical concerns and ensuring sustainable AI applications are essential for the responsible deployment of AI in climate forecasting.

Using machine learning to forecast clean energy, commodities, green bonds and ESG index prices: How important is financial stress?

PurposeThis study investigates clean energy, commodities, green bonds and environmental, social and governance (ESG) index prices forecasting and assesses the predictive performance of various factors on these asset prices, used for the development of a robust forecasting support decision model using machine learning (ML) techniques. More specifically, we explore the impact of the financial stress on forecasting price.Design/methodology/approachWe utilize feature selection techniques to evaluate the predictive efficacy of various factors on asset prices. Moreover, we have developed a forecasting model for these asset prices by assessing the accuracy of two ML models: specifically, the deep learning long short-term memory (LSTM) neural networks and the extreme gradient boosting (XGBoost) model. To check the robustness of the study results, the authors referred to bootstrap linear regression as an alternative traditional method for forecasting green asset prices.FindingsThe results highlight the significance of financial stress in enhancing price forecast accuracy, with the financial stress index (FSI) and panic index (PI) emerging as primary determinants. In terms of the forecasting model's accuracy, our analysis reveals that the LSTM outperformed the XGBoost model, establishing itself as the most efficient algorithm among the two tested.Practical implicationsThis research enhances comprehension, which is valuable for both investors and policymakers seeking improved price forecasting through the utilization of a predictive model.Originality/valueTo the authors' best knowledge, this marks the inaugural attempt to construct a multivariate forecasting model. Indeed, the development of a robust forecasting model utilizing ML techniques provides practical value as a decision support tool for shaping investment strategies.

Construction and Comparison of Machine-Learning Forecast Models of Albacore Thunnus alalunga Fishing Grounds in the South Pacific Ocean

The traditional methods for predicting the distribution of albacore (Thunnus alalunga) fishing grounds have low performance and accuracy. Uneven sampling can result in unreasonable evaluation indicators. To address these issues, three methods, equi-frequency, K-means clustering algorithm, and 1-R split, were applied to discretize the catch per unit effort (CPUE) of albacore in the South Pacific from 2016 to 2021 and partition the fishing grounds into abundance levels. Eight machine learning models were used to predict the fishing grounds. In addition to the traditional evaluation index based on confusion matrix, top-k index was also used to evaluate the accuracy of fishery abundance predictions. The results showed that (1) When sampling is unbalanced, the reported accuracy does not fully represent the actual performance of the model in predicting the abundance of albacore in the fishing ground. F1 value can be used as the index of the model effect and stability. (2) In binary classification, the quartile stacking algorithm has the best stacking performance, with F1 0.89. (3) The top-1 prediction accuracy of three-category fishery forecasting is the highest at 0.74, and the top-1 prediction accuracy of five-category fishery forecasting is the highest at 0.54. (4) The top-k accuracy of classification of fisheries with multiple abundance using K-means is significantly better than that of equal frequency discretization (p < 0.001). The top-k evaluation index was used to predict the fishing grounds of albacore across multiple abundance levels for the first time in this study, which is significant for pioneering a new method for this application and which provides a demonstration of the development of artificial intelligence techniques for fisheries in the future.

Book Value Forecast Quality: Financial Versus Nonfinancial Firms

ABSTRACTThis study investigates the quality and usefulness of analysts’ book value forecasts for financial firms relative to nonfinancial firms. We examine whether forecast properties (accuracy, bias, and dispersion) of book value forecasts and the market reaction to unexpected book value surprises differ between financial and nonfinancial firms. We find that book value forecasts are more accurate, less biased, and less dispersed for financial firms compared to nonfinancial firms. In addition, we show that these results are not simply a result of differences in earnings per share (EPS) forecast quality. We also provide evidence that book value forecasts are incrementally helpful to analysts in forecasting earnings for financial firms. In addition, we find evidence that the stock price drift following an earnings announcement is more pronounced when book value per share (BPS) surprises are considered in addition to EPS surprises. However, this finding only holds for financial firms. We find that this delayed reaction can form the basis for a profitable trading strategy.

Using dynamic semantic structure of news flow to enhance financial forecasting: a twelve-year study on twitter news channels

AbstractThis research holds significance for advancing financial forecasting methodologies by shifting the focus from traditional sentiment analysis of individual tweets to exploring intricate semantic relationships within news tweets from top-followed news channels on Twitter. Addressing a notable research gap in financial forecasting, often dominated by sentiment analysis, our study endeavors to fill the void left by the underexplored intricate relationships within news entities and their dynamic semantic evolution. Motivated by the inherent challenges in predicting the random walk behavior of stock prices, we contend that incorporating longitudinal data derived from the semantic relationships between news entities can enhance the accuracy of stock market forecasts. The study pioneers a twelve-year exploration, encompassing data from 55 leading news channels on Twitter, boasting a collective following of 714 million users. The approach employs natural language processing (NLP) to extract two million unique entities, whose semantics are analyzed through complex network analysis, laying the foundation for the forecasting model. Finally, this research introduces a model linked to the dynamic semantic structure of news flow. The predictive model considers the impact of exogenous variables influenced by the evolving relationships among news entities. The results offer a proof of concept, highlighting the potential of utilizing dynamic semantic relationships among news entities for financial prediction. On average, the model demonstrates an improvement in accuracy of 40.3% across ten different stock price predictions. These findings are expounded through relevant theories, offering a theoretical foundation for observed patterns and indicating a promising direction for future research in this domain.

MLFGCN: short-term residential load forecasting via graph attention temporal convolution network.

Residential load forecasting is a challenging task due to the random fluctuations caused by complex correlations and individual differences. The existing short-term load forecasting models usually introduce external influencing factors such as climate and date. However, these additional information not only bring computational burden to the model, but also have uncertainty. To address these issues, we propose a novel multi-level feature fusion model based on graph attention temporal convolutional network (MLFGCN) for short-term residential load forecasting. The proposed MLFGCN model fully considers the potential long-term dependencies in a single load series and the correlations between multiple load series, and does not require any additional information to be added. Temporal convolutional network (TCN) with gating mechanism is introduced to learn potential long-term dependencies in the original load series. In addition, we design two graph attentive convolutional modules to capture potential multi-level dependencies in load data. Finally, the outputs of each module are fused through an information fusion layer to obtain the highly accurate forecasting results. We conduct validation experiments on two real-world datasets. The results show that the proposed MLFGCN model achieves 0.25, 7.58% and 0.50 for MAE, MAPE and RMSE, respectively. These values are significantly better than those of baseline models. The MLFGCN algorithm proposed in this paper can significantly improve the accuracy of short-term residential load forecasting. This is achieved through high-quality feature reconstruction, comprehensive information graph construction and spatiotemporal features capture.

Predictive performance of denoising algorithms in S&P 500 and Bitcoin returns

This paper aims to investigate the effectiveness of both the Hankel matrix and the Wavelet denoise methods to address the issue of denoising volatility in datasets, specifically squared returns, and to enhance the accuracy of post-sample forecasting. When converting data from levels to volatility, noise levels are amplified, and the forecasting models suffer from structural bias, leading to reduced accuracy gains. The main challenge in denoising operations is preserving systemic impulses while extracting sporadic and unpredictable components. While conventional denoising algorithms struggle to distinguish systemic patterns from noise, the Hankel matrix approach surpasses them in performance, resulting in improved predictive accuracy. Additionally, this paper explores the Wavelet denoise method, which is found to be less effective than the Hankel matrix approach for GARCH models. The empirical work showcases the denoising process and the accuracy improvements achieved in various datasets for GARCH models. By implementing the Hankel matrix approach, simulations reveal a notable accuracy gain of over 10%.