Effective Forecasting Methods Research Articles

A novel hybrid forecasting model with feature selection and deep learning for wind speed research

AbstractAccurate wind speed prediction is of great importance for the operation of wind farms, and extensive efforts have been made to develop effective forecasting methods in this regard. However, the feature selection of data input as well as optimization of deep learning models have received comparatively less attention, leading to unreliable forecasting results. This research proposes a novel hybrid model that integrates data preprocessing, feature selection, and optimized forecasting for improved wind speed prediction. Specifically, a powerful preprocessing technique is utilized to reduce data noise disturbances, while an innovative two‐stage feature selection is designed to achieve the optimal input data format for forecasting purposes. Moreover, a hybrid forecasting module based on long‐short term memory, which is optimized by the Bayesian optimization algorithm, has been developed to enhance the efficiency and reliability of the model. The empirical study used 10‐min interval wind speed data of four seasons for presentation and evaluation results demonstrated its superior performance in effectively learning the volatility and irregularity features of wind speed series, which established a solid foundation for practical applications in wind power systems.

Performance evaluation of technical indicators for forecasting the moroccan stock index using deep learning

<span>Navigating the complex terrain of financial markets requires accurate forecasting tools, underscoring the need for effective forecasting methods to assist investors and policymakers alike. This paper explores deep learning techniques for forecasting the Moroccan all shares index (MASI), a prominent indicator of the Moroccan stock market. The study aims to evaluate the performance of technical indicators in enhancing the accuracy of MASI predictions. A comprehensive dataset of daily closing prices of the MASI index is collected and 26 technical indicators are computed from the historical price data. Deep learning models based on artificial neural networks (ANNs) are trained and optimized using the dataset. The performance of the models is evaluated using standard metrics such as mean absolute error (MAE), mean square error (MSE), mean absolute percentage error (MAPE), and root mean square error (RMSE). Additionally, feature selection techniques are employed to identify the subset of technical indicators that contribute most significantly to the prediction accuracy. The findings provide insights into the effectiveness of deep learning models and the impact of technical indicators on MASI prediction accuracy. This research has important implications for investors, financial analysts, and policymakers, enhancing investment strategies and risk management approaches.</span>

Predicting Traffic Casualties Using Support Vector Machines with Heuristic Algorithms: A Study Based on Collision Data of Urban Roads

Traffic accidents on urban roads are a major cause of death despite the development of traffic safety measures. However, the prediction of casualties in urban road traffic accidents has not been deeply explored in previous research. Effective forecasting methods for the casualties of traffic accidents can improve the manner of traffic accident warnings, further avoiding unnecessary loss. This paper provides a practicable model for traffic forecast problems, in which ten variables, including time characteristics, weather factors, accident types, collision characteristics, and road environment conditions, were selected as independent factors. A mixed-support vector machine (SVM) with a genetic algorithm (GA), sparrow search algorithm (SSA), grey wolf optimizer algorithm (GWO) and particle swarm optimization algorithm (PSO) separately are proposed to predict the casualties of collisions. Grounded on 4285 valid urban road traffic collisions, the computing results show that the SSA-SVM performs effectively in the casualties forecast compared with the GWO-SVM, GA-SVM and PSO-SVM.

Разработка гибридной модели прогнозирования потребления электрической энергии для горно-металлургического предприятия

Introduction. Reducing the energy intensity of production is one of the directions of development of enterprises, regions, states, as it affects the level of economic development and environmental well-being. One of the methods that contribute to the rationalization of the use of electric energy is forecasting. The purpose of the study is to develop a model for forecasting electricity consumption, which allows obtaining a reliable forecast of electricity consumption. To achieve this goal, an analysis of literary sources reflecting the current trends of predictive analytics in solving the problem of forecasting electricity consumption was carried out. Materials and methods of research. Singular spectral analysis and deep learning algorithms (multilayer perceptron) were used in the work. At the initial stage of the study, a model of a fully connected neural network (multilayer perceptron) was built, the prediction accuracy of which was more than 94%. Then the algorithm of singular spectral analysis was implemented, and the initial time series of power consumption was decomposed into additive components. The resulting decompositions were used to create a hybrid forecasting model. Research results. A number of experimental studies have been conducted that meet the goal of developing models that most accurately approximate experimental data. The result of the study are two forecasting models: an artificial neural network model and a hybrid model that includes a singular spectral analysis to decompose a number of power consumption into trend and harmonic components and an artificial neural network to predict them. Discussion. With the help of the hybrid model, an accurate result was obtained – the average absolute percentage error for the year is 4.04%. Forecasting was carried out for a year (by months) using data from the previous two years as a training subset. The neural network model provides less accuracy in predicting power consumption over 12 months compared to the hybrid model. Conclusion. Thus, the use of singular spectral analysis made it possible to reduce the magnitude of the forecast error by almost 1%, which confirms the effectiveness of the method and the relevance of the development of hybrid forecasting models. Resume. As a result of the conducted research, it was revealed that one of the most popular and effective forecasting methods is singular spectral analysis and the use of various deep learning algorithms. The developed hybrid model of electric energy consumption for a mining and metallurgical enterprise allows you to build a forecast of electricity consumption, the accuracy of which is more than 95%. Suggestions for practical application and directions for future research. The research results can be useful in the process of production management, since an accurate forecast of electricity consumption is important when planning production, scheduling equipment repairs and maintenance of the electrical network. The prospect of further research is to increase the forecasting horizon, the volume of the initial sample and taking into account additional factors characterizing the technological process.

Multi-step-ahead solar output time series prediction with gate recurrent unit neural network using data decomposition and cooperation search algorithm

As the solar energy develops sharply in recent years, accurate solar output forecasting is becoming one of the most important and challenging problems in modern power system. For enhancing the prediction accuracy of solar output, this research proposes an effective forecasting method using the famous compete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), gated recurrent unit (GRU) and cooperation search algorithm (CSA). The proposed methodology is composed of three important stages: firstly, the solar output signal is divided into a set of relatively simple subcomponents with obvious frequency differences via the CEEMDAN method; secondary, the GRU model is used to individually forecast each subcomponent while the CSA method is used to optimize the GRU parameters and enhance the forecasting ability; finally, the simulation values of all constructed models are added to obtain the corresponding forecasting results. The developed model takes advantages of the data decomposition technique and advanced machine learning to identify the suitable dependence relationship and network topology structures. Extensive experiments indicate that the developed model can yield accurate forecasting results for solar outputs in comparison with several traditional forecasting methods with respect to different evaluation criteria. Thus, an effective framework combining the signal decomposition technique and evolutionary method into machine learning model is presented for solar output forecasting.

Quantitative Analysis of Anesthesia Recovery Time by Machine Learning Prediction Models

It is significant for anesthesiologists to have a precise grasp of the recovery time of the patient after anesthesia. Accurate prediction of anesthesia recovery time can support anesthesiologist decision-making during surgery to help reduce the risk of surgery in patients. However, effective models are not proposed to solve this problem for anesthesiologists. In this paper, we seek to find effective forecasting methods. First, we collect 1824 patient anesthesia data from the eye center and then performed data preprocessing. We extracted 85 variables to predict recovery time from anesthesia. Second, we extract anesthesia information between variables for prediction using machine learning methods, including Bayesian ridge, lightGBM, random forest, support vector regression, and extreme gradient boosting. We also design simple deep learning models as prediction models, including linear residual neural networks and jumping knowledge linear neural networks. Lastly, we perform a comparative experiment of the above methods on the dataset. The experiment demonstrates that the machine learning method performs better than the deep learning model mentioned above on a small number of samples. We find random forest and XGBoost are more efficient than other methods to extract information between variables on postoperative anesthesia recovery time.

Model for Determining the Probability of Airline Bankruptcy

The present study is aimed at developing a preliminary model for assessing the financial condition and forecasting the probability of bankruptcy of the airline. The goal is using effective forecasting methods to develop a model for predicting the probability of bankruptcy of airlines in the conditions of non-stationary development of the Russian economy. In this paper the financial resources of mainly large Russian airlines and their patterns are considered, using economic and mathematical forecasting methods and general methods of scientific cognition in various aspects. The authors substantiate the existence of patterns among the financial indicators of Russian airlines of bankrupt and non-bankrupt, and also propose schemes for constructing an economic and mathematical model and identify median values of financial indicators. This model for preliminary detection of airline bankruptcies can be used taking into account a limited set of financial indicators due to the ease of implementation and the availability of most data in the public domain. The authors of the work investigated the methods of predicting bankruptcy of enterprises, with an emphasis on airlines, and made a comparative analysis of financial indicators according to the financial statements of Russian airlines.

Climate Change Forecasting Using Machine Learning SARIMA Model

Every country's population will have to deal with the effects of climate change. The meteorological department needs to implement effective forecasting methods to deal with climate changes. Accurate temperature forecasts help in protecting people and property is an essential aspect of government, business, and the general public planning. Early predictions help farmers and industrialists to make approaches and store crops more effectively. When the climate continuously changes, it is not easy to make accurate predictions for the meteorological department and government authorities. Artificial intelligence (AI) algorithms have stimulated improvements in various fields. Machine learning (ML) may find teleconnections where complicated feedbacks make it challenging to determine how proposed work from a straightforward analysis and observations. Our proposed research uses the machine learning algorithm, SARIMA Model, to comprehend and utilize existing datasets and simulations.

Time series research in the non-ferrous scrap recycling sector with identification of trend-seasonal components by Chetverikov method

The article is devoted to identifying of trend-seasonal components using iterative methods of the non-ferrous scrap metal collection filtering. According to the study objectives and the results of the time series characteristics analysis, the Chetverikov method implementation as a method for data filtering is described stepwise. Calculations and construction of diagrams were carried out in MS Excel. The study result is a visualized stepwise filtering of the trend-seasonal time series of the non-ferrous scrap mass in two positions - copper and brass, as well as the values of the seasonal component and specific seasonality indices calculated using the Chetverikov method, which makes it possible to study further the use and predicting of the non-ferrous metal scrap collection rate in the region, to select effective forecasting methods and models.

Improving Stock Price Prediction Using Combining Forecasts Methods

This study presents an outcome of pursuing better and effective forecasting methods. The study primarily focuses on the effective use of divide-and-conquer strategy with Empirical Mode Decomposition or briefly EMD algorithm. We used two different statistical methods to forecast the high-frequency EMD components and the low-frequency EMD components. With two statistical forecasting methods, ARIMA (Autoregressive Integrated Moving Average) and EWMA (Exponentially Weighted Moving Average), we investigated two possible and potential hybrid methods: EMD-ARIMA-EWMA, EMD-EWMA-ARIMA based on high and low-frequency components. We experimented with these methods and compared their empirical results with four other forecasting methods using five stock market daily closing prices from the S&P/TSX 60 Index of Toronto Stock Exchange. This study found better forecasting accuracy from EMD-ARIMA-EWMA than ARIMA, EWMA base methods and EMD-ARIMA as well as EMD-EWMA hybrid methods. Therefore, we believe frequency-based effective method selection in EMD-based hybridization deserves more research investigation for better forecasting accuracy.

A Meta-learning Based Multimodal Neural Network for Multistep Ahead Battery Thermal Runaway Forecasting

An effective forecast method to trigger Thermal Runaway (TR) warning in an early stage is essential for monitoring battery safety. In this article, we propose a novel data-driven approach to perform multistep ahead forecast accurately for battery TR state at cell-level. We formulate this forecasting task as an imbalance data classification task and propose meta thermal runaway forecasting neural network (Meta-TRFNN) to solve it. Essentially, we exploit high-dimensional thermal images along with low-dimensional temperature and voltage data to capture a more representative thermal profile. Moreover, we adapt a meta-learning framework to handle the data deficiency problem. We evaluate Meta-TRFNN on simulated samples and also explore its applicability in the real world with real samples. Although this classification task is highly imbalanced, Meta-TRFNN is still proven effective with limited historical information. Our further comparison experiments not only demonstrate the forecasting ability of Meta-TRFNN, but also validate the benefit of involving high-dimensional thermal images and the efficacy of meta-learning framework.

Real-Time Data and Flood Forecasting in Tagus Basin. A Case Study: Rosarito and El Burguillo Reservoirs from 8th to 12th March, 2018

The hydrological regime of the Iberian Peninsula is characterized by its extreme irregularity, including its propensity for periodic floods, which cause severe floods. The development of suitable cartographies and hydraulic models (HEC-RAS, IBER, etc.) allows for defining, with sufficient precision, the areas flooded by a determined return period, and for elaborating maps of danger and areas at risk of flooding, making it possible to adopt the corresponding preventive measures of spatial planning. These preventive measures do not avoid the need for contingent plans, such as the Civil Flooding Protection Plans. Many of the Peninsula’s watercourses and rivers are regulated by reservoirs built to ensure water supply and to smooth floods by releasing water in extreme hydrological climates. Hydrological modeling tools (rain/run-off) and Decision Support Systems DSS have been developed for the optimal operation of these dams in flood situations. The objective of the article is to study and prove the effectiveness of the integrated data provision in real time, while the event occurs—a circumstance that was not possible from the limited available meteorological stations available from Official Weather Services. The development of the Automatic Hydrological Information System (SAIH) in the Spanish River Basin Authorities (Confederaciones Hidrográficas), which includes a dense network of thermo-pluviometric stations and rain-river flow gauges, has allowed for new perspectives in order to realize an effective forecast method of flows during episodes of extreme precipitation. In this article, we will describe the integration of a hydrological modeling system, developed by the Confederación Hidrográfica del Tajo (River Authority), to meet the described objectives; the results of this methodology are novel. This allows for the processing of the 15-minute data provided, including the simulation of the snow accumulation/melting processes and the forecast of inflows to the reservoir to help in the establishment of safeguards and preventive waterflow releases. Finally, the methodology described is shown in a real case of study at Rosarito and El Burguillo Reservoirs.

A novel competitive swarm optimized RBF neural network model for short-term solar power generation forecasting

Solar power is an important renewable energy resource and acts as a major contributor to replacing fossil fuel generators and reducing carbon emissions. However, the intermittent power output due to the uncertain solar irradiance significantly challenges the economic integrations of solar generation within the existing power system, which calls for effective forecasting methods to improve the solar prediction accuracy. In this paper, a novel improved radial basis function neural network model is proposed and applied in forecasting the short-term solar power generation. A recent proposed meta-heuristic approach named competitive swarm optimization is adopted to train the non-linear and linear parameters of the radial basis function neural network model. The proposed model has been validated in nonlinear benchmark functions and then employed in forecasting the solar power generation of a real-world case study in the Netherlands. Numerical results demonstrate that the proposed competitive swarm optimized radial basis function neural network model could obtain higher accuracy compared to other counterparts and thus provides a useful tool for solar power forecasting.

Optimal Scheduling Method of Controllable Loads in Smart Home Considering Re-Forecast and Re-Plan for Uncertainties

Renewable energies (REs) such as photovoltaic generation (PV) have been gaining attention in distribution systems. Recently, houses with PV and battery systems, as well as electric vehicles (EV) are expected to contribute to not only the suppression of global warming but also reducing electricity bill on the consumer side. However, there are numerous challenges with the introduction of REs at the demand side such as the actual output of REs often deviating from the forecasted output, which causes fluctuation of the power flow and this is challenging for the distribution or transmission system operator. For this challenge, it is expected that smart grid technology using controllable loads such as a fixed battery or EV battery, can suppress fluctuation of power flow. This paper presents a decision method of optimal scheduling of controllable loads to suppress the fluctuation of power flow by PV output in the smart home. An optimization method to cope with uncertainties such as variability of PV power and effective forecasting methods are considered in the proposed scheme. In order to decrease the expected operational cost and to validate the robustness for the uncertainty’s optimization approach, statistical analysis is executed for the optimal scheduling scheme. From the optimization results, the proposed methodology suppressed the fluctuation of power flow in the smart home and also minimized the consumer operational cost.

Multi-Agent-Based Forecast Update Methods for Profit Enhancement of Intermittent Distributed Generators in a Smart Microgrid

Uncertainty in generation from intermittent sources makes a strong case to have effective forecasting methods. However, errors in forecast lead to losses to the distributed generation (DG) owners. In this article, multi-agent-based forecast update methods are proposed which minimize the forecast errors. The effectiveness of the proposed methods in enhancing the profit of intermittent generators and microgrid operational cost is analyzed using a microgrid with two scenarios, namely simple ownership and multiple ownership. A modified IEEE 13 bus system is used as the case study system and the system simulation for the microgrid is performed on the OpenDSS platform and the proposed multi-agent system is developed using JAVA Agent DEvelopment (JADE) framework. From the simulation results, the proposed approaches are found effective in increasing profit margins for the investors or owners of the DGs.

Experimental and theoretical studies of the dynamic behavior of a spiral-groove dry gas seal at high-speeds

Abstract Because of great axial vibration of the rotor system, dry gas seal has a big film thickness disturbance, which may cause contact rubbing of seal faces or excessive leakage. Focusing on this problem, experimental studies on transient-state film thickness and leakage rate of a spiral groove dry gas seal at high-speeds are carried out under different spring pressures and different spiral groove depths. And the experimental results of film thickness disturbance and average leakage rate are predicted by perturbation method and gas lubrication theory. Results show that the theoretical results of variation trends of film thickness disturbance amplitudes coincide well with the experimental data, and it also represents that the perturbation method is an effective forecasting method in dynamic analysis.

Integrating metaheuristics and artificial neural network for weather forecasting

Over the years, researchers have been analysing to forecast the weather as precisely as possible in order to provide better living conditions. Nevertheless, there is no consensus on the effective weather forecasting methods and therefore, research on providing applicable and effective forecasting methods has been continued. In this study, artificial neural networks (ANNs) are integrated with two metaheuristic methods including genetic algorithm (GA) and harmony search (HS) to determine the most relevant input variables and to search the most appropriate number of hidden neurons. The proposed forecasting methods are implemented for six different cities of Turkey that are selected according to Aydeniz's climate classification. The results of the graphical analysis and performance measures show that daily mean temperature forecasting is improved by GA-ANN and HS-ANN methods due to the ability to capture the advantages of metaheuristic and ANN simultaneously.

Forecasting Demand in Supply Chain Using Machine Learning Algorithms

Managing inventory in a multi-level supply chain structure is a difficult task for big retail stores as it is particularly complex to predict demand for the majority of the items. This paper aims to highlight the potential of machine learning approaches as effective forecasting methods for predicting customer demand at the first level of organization of a supply chain where products are presented and sold to customers. For this purpose, we utilize Artificial Neural Networks (ANNs) trained with an effective second order algorithm, and Support Vector Machines (SVMs) for regression. We evaluated the effectiveness of the proposed approach using public data from the Netflix movie rental online DVD store in order to predict the demand for movie rentals during an especially critical for sales season, which is the Christmas holiday season. In our analysis we also integrated data from two other sources of information, namely an aggregator for movie reviews (Rotten Tomatoes), and a movie oriented social network (Flixster). Consequently, the approach presented in this paper combines the integration of data from various sources of information and the power of advanced machine learning algorithms for lowering the uncertainty barrier in forecasting supply chain demand.

The Realised–Implied Volatility Relationship: Recent Empirical Evidence from FTSE‐100 Stocks

ABSTRACTThis paper examines the long‐run relationship between implied and realised volatility for a sample of 16 FTSE‐100 stocks. We find strong evidence of long‐memory, fractional integration in equity volatility and show that this long‐memory characteristic is not an outcome of structural breaks experienced during the sample period. Fractional cointegration between the implied and realised volatility is shown using recently developed rank cointegration tests by Robinson and Yajima (2002). The predictive ability of individual equity options is also examined and composite implied volatility estimates are shown to contain information on future idiosyncratic or stock‐specific risk that is not captured using popular statistical approaches. Implied volatilities on individual UK equities are thus closely related to realised volatility and are an effective forecasting method particularly over medium forecasting horizons. Copyright © 2011 John Wiley & Sons, Ltd.

Variations in the lower level of the PBL associated with the Yellow Sea fog-new observations by L-band radar

The Chinese east coastal areas and marginal seas are foggy regions. The development of effective forecasting methods rests upon a comprehensive knowledge of the fog phenomena. This study provides new observations associated with the sea fogs over the northwestern Yellow Sea by means of L-band radar soundings with a high vertical resolution of 30 m. The monthly temperature lapse rate, the Richardson Numbers, and the humidity show obvious seasonal variations in the lower level of the planetary boundary layer (PBL) that are related to the onset, peak and end of the Yellow Sea fog season. The typical pattern of stratification for the sea fog season in the northwestern Yellow Sea is that a stable layer of about 400 m thick caps a 150 m conditionally unstable layer. Besides, the differences between fogs and stratus clouds in terms of humidity, turbulence and temperature are analyzed, which is of significance for sea fog forecast and detection by satellites. The thickness of the sea fogs varies in different stages of the fog season, and is associated with the temperature inversion. The numerical simulation proves that the seasonal variations obtained by the radar well represent the situations over the Yellow Sea.