Short-term Forecasting Research Articles

Review and Intercomparison of Machine Learning Applications for Short-term Flood Forecasting

Abstract Among natural hazards, floods pose the greatest threat to lives and livelihoods. To reduce flood impacts, short-term flood forecasting can contribute to early warnings that provide communities with time to react. This manuscript explores how machine learning (ML) can support short-term flood forecasting. Using two methods [strengths, weaknesses, opportunities, and threats (SWOT) and comparative performance analysis] for different forecast lead times (1–6, 6–12, 12–24, and 24–48 h), we evaluate the performance of machine learning models in 94 journal papers from 2001 to 2023. SWOT reveals that the best short-term flood forecasting was produced by hybrid, random forest (RF), long short-term memory (LSTM), artificial neural network (ANN), and adaptive neuro-fuzzy inference system (ANFIS) approaches. The comparative performance analysis, meanwhile, favors convolutional neural network, ANFIS, multilayer perceptron, k-nearest neighbors algorithm (KNN), hybrid, LSTM, ANN, and support vector machine (SVM) at 1–6 h; hybrid, ANFIS, ANN, and LSTM at 6–12 h; SVM, hybrid, and RF at 12–24 h; and hybrid and RF at 24–48 h. In general, hybrid approaches consistently perform well across all lead times. Trends such as hybridization, model selection, input data selection, and decomposition seem to improve the accuracy of models. Furthermore, effective stand-alone ML models such as ANN, SVM, RF, genetic algorithm, KNN, and LSTM, provide better outcomes through hybridization with other ML models. By including different machine learning models and parameters such as environmental, socio-economical, and climatic parameters, the hybrid system can produce more accurate flood forecasting, making it more effective for early warning operational purposes.

Short-term Load Forecasting Based on K-medoids Clustering and XGBTide Model

Introduction: Electric power load is significantly influenced by weather conditions, making accurate load prediction under varying weather scenarios essential for effective planning and stable operation of the power system. This paper introduces a short-term load forecasting method that combines k-Medoids clustering and XGB-TiDE. Initially, k-Medoids clusters the original load data into categories such as sunny, high temperature, and rain/snow days. Subsequently, XGBoost identifies critical features within these subsequences. The combined forecast model, XGB-TiDE, is then tailored for each subsequence. Here, the TiDE model's predictions are refined point-by-point using the XGBoost results to derive the final short-term load forecasts. An empirical analysis using real power load data from a specific region demonstrates that our proposed model achieves superior accuracy, especially under extreme weather conditions such as high temperatures and precipitation. Background: Electric power load is significantly influenced by weather conditions, making accurate load prediction under varying weather scenarios essential for effective planning and stable operation of the power system. Objective: Addressing the limitations of short-term load forecasting under extreme weather conditions, this paper introduces a novel approach that leverages k-Medoids clustering and the XGBTiDE model to enhance forecasting accuracy. This method strategically segments power load data into meaningful clusters before applying the robust predictive capabilities of XGB-TiDE, aiming for a significant improvement in forecast precision. Method: Initially, k-Medoids clusters the original load data into categories such as sunny, high temperature, and rain/snow days. Subsequently, XGBoost identifies critical features within these subsequences. The combined forecast model, XGB-TiDE, is then tailored for each subsequence. Here, the TiDE model's predictions are refined point-by-point using the XGBoost results to derive the final short-term load forecasts. Results: The model presented in this study demonstrates outstanding performance across all weather conditions, consistently achieving lower mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) compared to other models. For instance, on a sunny day, this model records an MAE of 8.49, RMSE of 10.29, and MAPE of 2.55%, markedly surpassing the Autoformer, which shows an MAE of 18.29, RMSE of 22.33, and MAPE of 5.50%. These results underscore the superior accuracy of our proposed forecasting approach. Conclusion: An empirical analysis using real power load data from a specific region demonstrates that our proposed model achieves superior accuracy, especially under extreme weather conditions such as high temperatures and precipitation.

THE COMPARISON OF ARIMA AND RNN FOR FORECASTING GOLD FUTURES CLOSING PRICES

In the financial markets, accurately forecasting the closing prices of gold futures is crucial for investors and analysts. Traditional methods like ARIMA (Autoregressive Integrated Moving Average) have been widely used for this purpose, particularly for their effectiveness in short-term stable data forecasting. However, the inherent complexity and dynamic nature of financial data, coupled with trends and seasonal patterns, present significant challenges for long-term forecasting with ARIMA. Conversely, advanced methods such as Recurrent Neural Networks (RNN) have shown promise in handling these complexities and providing reliable long-term forecasts. This research seeks to evaluate and compare the performance of ARIMA and RNN in forecasting daily gold futures closing prices using forecast accuracy tests namely RMSE and MAPE, aiming to identify the optimal method that balances accuracy, stability, and adaptability to trends and seasonal variations in the financial market. The daily data for this analysis is sourced from Investing.com (https://www.investing.com).

Do economic shocks matter to U.S. farm income?

ABSTRACT Analysing the effect of economic shocks on farm profitability is crucial for determining the U.S. farm economy’s vulnerability to macroeconomic changes. Using a structural vector autoregressive model and over five decades of data, we examine the impact and importance of external economic shocks to the U.S. farm economy. Specifically, we measure the short-term and long-term volatility spillover effects on U.S. farm income from shocks in energy markets, exchange rates, and agricultural markets. We find that volatility in exchange rates significantly affects US net farm income fluctuations in the short term. Results also suggest that shocks in the exchange rate are more important than other economic shocks, explaining about 47% of the short-run fluctuation in U.S. net farm income. These results highlight the importance of accounting for macroeconomic shocks to enhance the stability of the U.S. agricultural sector and to improve the accuracy of USDA short-term forecasts and long-term projections of farm income components.

Forecasting China’s Short-Term Energy Futures Price Using a Novel Secondary Decomposition-Optimized System

AbstractEnergy futures price forecasting is challenging due to the nonlinear and fluctuant characteristics. Existing literature mainly uses decomposition and ensemble method which neglects the intrinsic mode function obtained by the first decomposition could be irregular and thus reduces the prediction accuracy. To fill the research gap, a novel secondary decomposition-optimized-KELM-ensemble forecasting system is proposed to perform short-term forecasting in this study, which synthesizes two-stage data decomposition method, Sparrow search optimization algorithm, and extreme learning machine with kernel. We test the method with two energy futures prices in China, demonstrating that both one-day and three-day ahead forecasting results obtained are more accurate and stable compared to existing models in the literature, such as BPNN (improved by 58.42% on one-day ahead and 56.44% on three-day ahead by MAE) and KELM (improved by 56.40% on one-day ahead and 49.04% on three-day ahead by MAE). Therefore, the forecasting system introduced in this paper can provide useful implications for both policy makers and financial practitioners in the energy sector.

Projections of Public Spending on Pharmaceuticals: A Review of Methods.

Forecasting future public pharmaceutical expenditure is a challenge for healthcare payers, particularly owing to the unpredictability of new market introductions and their economic impact. No best-practice forecasting methods have been established so far. The literature distinguishes between the top-down approach, based on historical trends, and the bottom-up approach, using a combination of historical and horizon scanning data. The objective of this review is to describe the methods for projections of pharmaceutical expenditure that apply the "bottom-up" approach and to synthesize the knowledge of their predictive accuracy. Projections of public pharmaceutical expenditure applicable to Western economies including a comprehensive method description and published 2000-2024 were searched in scientific databases (MEDLINE, EMBASE, and EconLit) and in gray literature (websites of international health organizations and national healthcare authorities). The data sources, assumptions about the future market dynamics, analytical approaches, and the projection results are summarized. Twenty-four out of 3492 screened publications were included, associated with nine expenditure projection models. Four models were developed for all reimbursable drugs in the USA, the UK, the Stockholm region (Sweden), and seven European Union (EU) countries: France, Germany, Greece, Hungary, Poland, Portugal, and the UK, respectively. The other five models concerned specific groups of medicines: orphan drugs in Belgium, the Eurozone plus the UK, and Canada, respectively; psychotropic medications in the USA; and outpatient intravenous cancer medicines in the Province of Ontario (Canada). For trend analysis, drug coverage claims or sales data were used, applying linear and/or nonlinear models. The budget impact of new launches and patent expirations was estimated through (a form of) horizon scanning, i.e., a systematic monitoring of the pharmaceutical pipeline, with engagement of clinical expert judgment. Projections with a predictive time window greater than 3 years largely relied on previously observed trends to model new market introductions. Four models were validated through an ex post comparison of projected and observed expenditure. The absolute difference between the forecasted and actual percentual change in pharmaceutical expenditure was: 0.3% ("UK model"), 1.9% ("Stockholm model"), and 2% (nonfederal hospitals, "US model"). The "Ontario cancer drug model" overestimated the actual expenditure by 1%. Overall, the largest errors were attributable to new market launches and unforeseen policy reforms. Prediction accuracy decreased substantially for forecasts beyond 1 year in the future. For two not validated projections, a face validity check was feasible. One of the models forecasted a decrease in pharmaceutical expenditure from 2012 to 2016 in six European countries, contrasting with the currently available statistics. A 10-year projection of orphan drug expenditure underestimated the number of rare diseases treated in Europe by over 100%. Published projections of national pharmaceutical expenditure are scarce and marked by significant methodological variability. Short-term forecasts based on high-quality historical data and rigorous horizon scanning tend to be more accurate than long-term forecasts built on theoretical assumptions. The combination of mathematical algorithms and expert judgment should be further explored, to increase the accuracy and efficiency of pharmaceutical expenditure projections.

Hybrid CNN-Multivariate LSTM Model for Accurate Short-Term Electricity Price Forecasting and Energy System Optimization

Hybrid CNN-Multivariate LSTM Model for Accurate Short-Term Electricity Price Forecasting and Energy System Optimization

MIESTC: A Multivariable Spatio-Temporal Model for Accurate Short-Term Wind Speed Forecasting

Wind speed forecasting is an essential part of weather prediction, with significant value in economics, business, and management. Utilizing multiple meteorological variables can improve prediction accuracy, but existing methods face challenges such as mixing and noise due to variable differences, as well as difficulty in capturing complex spatio-temporal dependencies. To address these issues, this study introduces a novel short-term wind speed forecasting model named as MIESTC. The proposed model employs an independent encoder to extract features from each meteorological variable, mitigating the issues of noise that are caused by variable mixing. Then, a multivariate spatio-temporal correlation module is used to capture the global spatio-temporal dependencies between variables and model their interactions. Experimental results on the ERA5-LAND dataset show that, compared to the ConvLSTM, UNET, and SimVP models, the MIESTC model reduces RMSE by 14.60%, 8.64%, and 10.41%, respectively, for a 1 h prediction duration. For a 6 h prediction duration, the corresponding reductions are 13.91%, 8.20%, and 6.95%, validating its superior performance in short-term wind speed forecasting. Furthermore, an analysis of variable impacts reveals that U10, V10, and T2M play dominant roles in wind speed prediction, while TP exhibits a relatively lower impact, aligning with the results of the correlation analysis. These findings underscore the potential of MIESTC as an effective and reliable tool for short-term wind speed prediction.

A study on the runoff prediction mechanism of "water-soil-heat" in cold alpine regions with complex spatial distribution.

To address the challenge of runoff prediction in cold alpine regions with complex spatial distributions, this study proposes an integrated "Water-Soil-Hseat" framework for runoff modeling. This framework incorporates key factors such as precipitation, glacier meltwater, soil spatial distribution, and temperature-induced melt processes, providing a more comprehensive understanding of runoff generation mechanisms. Precipitation and glacier meltwater serve as the primary hydrological variables, while soil spatial distribution acts as an impact factor, and temperature-induced melt processes drive the runoff. The model uses these factors to dynamically simulate hydrological behavior under different topographical conditions. By combining a glacier module, a non-glacial area grid runoff generation model (GRGM), and an improved Transformer-based sequence model, the study develops a hybrid runoff prediction model-Glacier-GRGM-Transformer. The model is validated using daily runoff data from the Tailan River Basin from 2001 to 2020. The results show that the dynamic optimization of the Glacier-GRGM runoff generation model achieves a relative error of 6.79% and a coefficient of determination of 0.97, indicating that the model effectively simulates the runoff generation process of the watershed. The Glacier-GRGM-Transformer runoff prediction model demonstrates strong predictive capability in short-term forecasts, particularly within the first 4days, with NSE values consistently exceeding 0.7. For forecast periods longer than 2-3days, the model still provides reasonable forecasts, though with some decrease in accuracy, demonstrating its adaptability over longer timescales. Compared to the SRM and GRU-Transformer models, the Glacier-GRGM-Transformer model outperforms in both prediction accuracy and stability, further validating the adaptability and effectiveness of the "Water-Soil-Heat" framework in complex environments. This study not only effectively captures the spatial heterogeneity of runoff mechanisms but also enhances the model's ability to simulate and predict mixed runoff in complex environments by integrating both glacier and non-glacier hydrological modules. The research provides important scientific insights for hydrological simulation and forecasting in cold alpine regions.

Short-term load forecasting of electric vehicle charging stations based on adaptive VMD-BiLSTM

Short-term load forecasting of electric vehicle charging stations based on adaptive VMD-BiLSTM

Unified carbon emissions and market prices forecasts of the power grid

Carbon emissions and market prices forecasts of the power grid are of great importance for all electricity traders and consumers. Both forecasts enable flexible demand scheduling, ensuring sustainability and cost-efficiency. Many studies show the benefits of using both forecasts independently but not in combination, which remains an unexplored problem. The latest state-of-the-art techniques involve advanced statistical and machine learning algorithms leveraging seasonal patterns and exogenous forecasts. However, most of the reported studies deal only with problems of modeling and feature engineering, neglecting the forecast and model errors, which accumulate within the time-evolving power grid. This research aims to tackle these issues by introducing a versatile framework for short-term probabilistic forecasting of unified carbon emissions and market prices for electricity intra-day market participants. The approach utilizes the Hidden Markov Model for predictive estimation to determine the future energy mix of the country’s power grid. Furthermore, a novel optimization-based strategy, Moving Horizon Predictive Correction, is proposed to enhance the estimated energy mix performance, minimizing forecast and model errors. Subsequently, two separate recurrent neural networks are trained to provide probabilistic forecasts of carbon emissions and market prices, accounting for the stochastic dynamic of the power grid. A comparative analysis examines six case studies from various European countries and compares them with state-of-the-art forecasting methods. The results indicate that the proposed method can improve the qualitative measures by up to 53% for carbon emissions and up to 18% for market prices forecasts. Besides improving traditional point perditions, methods show significant increases in the quality of prediction intervals. Further application of the proposed forecasts is employed for a flexible 4–hour electricity consumption schedule. This showcases the usage of the proposed forecasts to find the best possible trade-off for low carbon emission, cost-effective electricity consumption time slots.

MPM: Multi Patterns Memory Model for Short-Term Time Series Forecasting

MPM: Multi Patterns Memory Model for Short-Term Time Series Forecasting

Multi-Task Graph Adaptive Learning for Multivariate Electricity Price Short-Term Forecasting in Australia's National Electricity Market

Multi-Task Graph Adaptive Learning for Multivariate Electricity Price Short-Term Forecasting in Australia's National Electricity Market

A Heterogeneity-Aware Adaptive Federated Learning Framework for Short-Term Forecasting in Electric IoT Systems

A Heterogeneity-Aware Adaptive Federated Learning Framework for Short-Term Forecasting in Electric IoT Systems

Research on short-term precipitation forecasting method based on CEEMDAN-GRU algorithm

Precipitation forecasting is vital for managing disasters, urban traffic, and agriculture. This study develops an improved model for short-term precipitation forecasting by combining Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Gated Recurrent Unit (GRU). Using precipitation data from January 1, 2019, to December 31, 2022, as a sample, the model capitalizes on CEEMDAN’s superior signal decomposition capabilities and GRU’s ability to capture nonlinear dynamic patterns in time series. To assess the model’s effectiveness, comparisons were conducted with 12 benchmark models, including CEEMDAN-LSTM, EMD-GRU, EMD-LSTM, BI-LSTM, GRU, LSTM, and TCN. The results demonstrate that the CEEMDAN-GRU model achieves higher accuracy and stability in short-term precipitation forecasting. Leveraging an Adam optimizer with adaptive learning rate reduction enhances convergence and ensures reliable predictions, achieving an R²of 0.7915, MAE of 0.05382, and MSE of 0.09081.

Flood Hazard Assessment Using Weather Radar Data in Athens, Greece

Weather radar plays a critical role in flash flood forecasting, providing an effective and comprehensive guide for the identification of possible flood-prone areas. However, the utilization of radar precipitation data remains limited in current research and applications, particularly in addressing flash flood hazards in complex environments such as in Athens, Greece. To address this gap, this study introduces the Gridded Flash Flood Guidance (GFFG) method, a short-term flash flood forecasting and warning technology based on radar precipitation and hydrological model coupling, and implements it in the region of Athens, Greece. The GFFG system improves upon the traditional flash flood guidance (FFG) concept by better integrating the weather radar dataset’s spatial and temporal flexibility, leading to increased resolution results. Results from six flood events underscore its ability to identify high-risk areas dynamically, with urban regions frequently flagged for flooding unless initial soil moisture conditions are low. Moreover, the sensitivity analysis of the system showed that the most crucial parameter apart from rainfall input is the soil moisture conditions, which define the amount of effective rainfall. This study highlights the significance of incorporating radar precipitation and real-time soil moisture assessments to improve flood prediction accuracy and provide valuable flood risk assessments.

Research on the Forecasting of Electricity Demand in India State Maharashtra

This study centers on forecasting the electricity demand in the Indian state of Maharashtra. In the context of rapid technological and economic development, the demand for basic raw materials and energy is on the rise. Moreover, the conflict between Russia and Ukraine has imposed supply constraints, making the prediction of electricity demand of utmost importance. Various approaches have been utilized in this field; however, existing studies often lack specificity and tend to be overly general. This research employs time series analysis, sourcing data from POSOCO. The ARIMA model is selected, and its parameters are determined through multiple rounds of testing. The results demonstrate that the ARIMA (1,0,1) model effectively fits the data, offering valuable insights. Overall, this study presents short-term forecasting and emphasizes the necessity of conducting specific and comprehensive analysis for the Intelligent Power System. Such analysis is crucial for optimizing power generation, distribution, and consumption, ensuring the stability and reliability of the power supply in Maharashtra. It also provides a foundation for decision-making in power system planning and management, enabling stakeholders to make informed choices regarding infrastructure investment and resource allocation. Additionally, the study's findings can contribute to the development of more efficient energy policies and strategies, promoting sustainable economic growth and social development in the region.

A multivariate soil temperature interval forecasting method for precision regulation of plant growth environment.

Foliage plants have strict requirements for their growing environment, and timely and accurate soil temperature forecasts are crucial for their growth and health. Soil temperature exhibits by its non-linear variations, time lags, and coupling with multiple variables, making precise short-term multi-step forecasts challenging. To address this issue, this study proposes a multivariate forecasting method suitable for soil temperature forecasting. Initially, the influence of various environmental factors on soil temperature is analyzed using the gradient boosting tree model, and key environmental factors are selected for multivariate forecasting. Concurrently, a point and interval forecasting model combining the Neural Hierarchical Interpolation for Time Series Forecasting (N-HiTS) and Gaussian likelihood function is proposed, providing stable soil temperature forecasting for the next 20 to 120 minutes. Finally, a multi-objective optimization algorithm is employed to search for optimal initial parameters to ensure the best performance of the forecasting model. Experiments have demonstrated that the proposed model outperforms common models in predictive performance. Compared to Long Short-Term Memory (LSTM) model, the proposed model reduces the Mean Absolute Error (MAE) for forecasting soil temperatures over the next 20, 60, and 120 minutes by 0.065, 0.138, and 0.125, respectively. Moreover, the model can output stable forecasting intervals, effectively mitigating the instability associated with multi-step point forecasts. This research provides a scientific method for precise regulation and disaster early warning in facility cultivation environments.

Comparing Time Series Predictions of COVID-19 Deaths Using SARIMAX, Neural Network, and XGBoost

Accurate and timely prediction of the development of the Covid-19 pandemic is critical to enabling adaptive responses and supporting better decision-making in pandemic management. This research aims to evaluate the effectiveness of three prediction models—SARIMAX, Neural Network, and XGBoost—in forecasting Covid-19-related deaths worldwide. Historical data from July 1 to August 15, 2020, was utilized to develop and test these models, with performance assessed using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). The findings reveal that the XGBoost model outperformed the other two, achieving the lowest MAE value of 347.87, demonstrating its superior accuracy for short-term trend forecasting. These results highlight the strengths and limitations of various forecasting methods, offering valuable insights for future research and practical applications in epidemic prediction. The implications of this research underscore the potential of advanced predictive models in enabling faster identification and response to health threats, thus contributing to more effective epidemic control and public health preparedness.

Forecasting the Number of Public Health Medical Beds in China: Based on ARIMA Model

This study employs the ARIMA model to forecast hospital bed availability in China from 1994 to 2023, with a focus on predicting bed numbers for the subsequent five years (20242028). The analysis encompasses two principal phases: the first phase involves forecasting future bed numbers based on historical data, and the second phase evaluates the accuracy of these forecasts by comparing them with actual data from 2021 to 2023. Using the ARIMA(0,2,0) model, the study demonstrates a robust and reliable forecast, revealing a steady increase in hospital bed counts. The accuracy of the model's predictions is high, confirming its effectiveness for short-term forecasting and underscoring the importance of proactive healthcare infrastructure planning. Additionally, the study highlights potential areas for future research, including the integration of additional variables, exploration of alternative modeling approaches, extension of forecast horizons, and sensitivity analyses to enhance predictive precision. These contributions aim to assist healthcare planners and policymakers in addressing the evolving needs of China's healthcare system and ensuring adequate medical service provision.