Search

Accurately forecasting energy metrics is essential for efficiently managing renewable energy generation. Given the high variability in load and renewable energy power output, this represents a crucial area of research in order to pave the way for increased adoption of low-carbon energy solutions. Whilst the impact of different neural network architectures and algorithmic approaches has been researched extensively, the impact of utilising additional weather variables in forecasts have received far less attention. This article demonstrates that weather variables can have a significant influence on energy forecasting and presents methodologies for using these variables within a long short-term memory (LSTM) architecture to achieve improvements in forecasting accuracy. Moreover, we introduce the use of the seasonal components of the target time series, as exogenous variables, that are also observed to increase accuracy. Load, solar and wind generation time series were forecast one hour ahead using an LSTM architecture. Time series data were collected in five Spanish cities and aggregated for analysis, alongside five exogenous weather variables, also recorded in Spain. A variety of LSTM architectures and hyperparameters were investigated. By tuning exogenous weather variables, a 33% decrease in mean squared error was observed for solar generation forecasting. A 22% decrease in mean absolute squared error (MASE), compared to 24-h ahead forecasts made by the Transmission Service Operator (TSO) in Spain, was also observed for solar generation. Compared to using the target variable in isolation, utilising exogenous weather variables decreased MASE by approximately 10%, 15% and 12% for load, solar and wind generation, respectively. By using the seasonal component of the target variables as an exogenous variable itself, we demonstrated decreases in MASE of 19%, 12% and 8% for load, solar and wind generation, respectively. These results emphasise the significant benefits of incorporating weather and seasonal components into energy-related time series forecasts. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

As CO2 emissions continue to rise, the problem of global warming is becoming increasingly serious. It is important to provide a robust management decision-making basis for the reductions of carbon emissions worldwide by predicting carbon emissions accurately. However, affected by various factors, the prediction of carbon emissions is challenging due to its nonlinear and nonstationary characteristics. Thus, we propose a combination forecast model, named CEEMDAN-GWO-SVR, which incorporates multiple features to predict trends in China's carbon emissions. First, the impact of online search attention and public health emergencies are considered in carbon emissions prediction. Since the impact of different variables on carbon emissions is lagged, the grey relational degree is used to identify the appropriate lag series. Second, irrelevant features are eliminated through RFECV. To address the issue of feature redundancy of online search attention, we propose a dimensionality reduction method based on keyword classification. Finally, to evaluate the features of the proposed framework, four evaluation indicators are tested in multiple machine learning models. The best-performed model (SVR) is optimized by CEEMDAN and GWO to enhance prediction accuracy. The empirical results indicate that the proposed framework maintains good performance in both multi-scenario and multi-step prediction. [ABSTRACT FROM AUTHOR]

Time series forecasting has a wide range of applications in various fields. To eliminate the need for time series data volume, a meta-learning-based few-shot time series forecasting method is proposed. This method uses a residual stack module as its backbone and connects the residuals forward and backward through a multilayer fully connected network so that the model and the meta-learning framework can be seamlessly combined. The Empirical knowledge of different time-sequence tasks is obtained through meta-training. To enable fast adaptation to new prediction tasks, a small meta-network is introduced to adaptively and dynamically generate the learning rate and weight decay coefficient of each step in the network. This method can use sequences of different data distribution characteristics for cross-task learning, and each training task only needs a small number of time series to achieve sequence prediction for the target task. The results show that compared with the two baselines, the proposed method has improved performance on 67.07% and 58.53% of the evaluated tasks. Thus, this method can effectively alleviate the problems caused by insufficient data during training and has broad application prospects in the field of time series. [ABSTRACT FROM AUTHOR]

In this paper, we present a new fuzzy symbolization technique for energy load forecasting with neural networks, FPLS-Sym. Symbolization techniques transform a numerical time series into a smaller string of symbols, providing a high-level representation of time series by combining segmentation, aggregation and discretization. The dimensional reduction obtained with symbolization can speed up substantially the time required to train neural networks, however, it can also lead to considerable information losses that could lead to a less accurate forecast. FPLS-Sym introduces the use of fuzzy logic in the discretization process, maintaining more information about each segment of the neural network at the expense of requiring more space in memory. Extensive experimentation was made to evaluate FPLS-Sym with various neural-network-based models, including different neural network architectures and activation functions. The evaluation was done with energy demand data from Spain taken from 2009 to 2019. Results show that FPLS-Sym provides better quality metrics than other symbolization techniques and outperforms the use of the standard numerical time series representation in both quality metrics and training time. [ABSTRACT FROM AUTHOR]

The adjusted precision of economic values is essential in the global economy. In recent years, researchers have increased their interest in making accurate predictions in this type of time series; one of the reasons is that the characteristics of this type of time series makes predicting a complicated task due to its non-linear nature. The evolution of artificial neural network models enables us to research the suitability of models generated for other purposes, applying their potential to time series prediction with promising results. Specifically, in this field, the application of transformer models is assuming an innovative approach with great results. To improve the performance of this type of networks, in this work, the empirical model decomposition (EMD) methodology was used as data preprocessing for prediction with a transformer type network. The results confirmed a better performance of this approach compared to networks widely used in this field, the bidirectional long short term memory (BiLSTM), and long short term memory (LSTM) networks using and without EMD preprocessing, as well as the comparison of a Transformer network without applying EMD to the data, with a lower error in all the error metrics used: The root mean square error (RMSE), the root mean square error (MSE), the mean absolute percentage error (MAPE), and the R-square (R²). Finding a model that provides results that improve the literature allows for a greater adjustment in the predictions with minimal preprocessing. [ABSTRACT FROM AUTHOR]

As a consequence of road accidents, around 1.3 million people die, and between 20 to 50 million have nonfatal injuries. Therefore, hospitals are receiving a high volume of patients in their urgent care, and a quick decision must be made regarding their treatment plans. At the admission stage, there is no information or probability about the patient's final result, regardless of if the patient will mostly die or be safely discharged from the hospital. To address this issue, this study proposed a machine learning-based framework that can predict the hospital disposition for trauma patients. The framework was developed to anticipate whether the patient would be safely discharged from the hospital or die based on a set of features collected at the admission time. In this study, the data used was collected from the King Abdulaziz Medical City (KAMC) in Riyadh, Saudi Arabia, and the performance of different machine learning algorithms was investigated, including eXtreme gradient boost (XGBoost), K-nearest neighbor, random forest, logistic regression, BRR, and support vector machine. Results show that the XGBoost algorithm demonstrated a high degree of detection and prediction accuracy for disposed-to-home patients; of the 6059 patients that were sent home, the XGBoost correctly predicted 5944 (98%) of the total. Finally, the developed framework could accurately predict hospital disposition for trauma patients with high accuracy and sensitivity levels. This system can benefit healthcare teams and insurance companies by providing them with a quick decision-making tool to determine the best treatment plan for patients. [ABSTRACT FROM AUTHOR]

Accurately predicting the future trend of a time series holds immense importance for decision-making and planning across various domains, including energy planning, weather forecasting, traffic warning, and other practical applications. Recently, deep learning methods based on transformers and time convolution networks (TCN) have achieved a surprising performance in long-term sequence prediction. However, the attention mechanism for calculating global correlation is highly complex, and TCN methods do not fully consider the characteristics of time-series data. To address these challenges, we introduce a new learning model named wavelet-based Fourier-enhanced network model decomposition (W-FENet). Specifically, we have used trend decomposition and wavelet transform to decompose the original data. This processed time-series data can then be more effectively analyzed by the model and mined for different components in the series, as well as capture the local details and overall trendiness of the series. An efficient feature extraction method, Fourier enhancement-based feature extraction (FEMEX), is introduced in our model. The mechanism converts time-domain information into frequency-domain information through a Fourier enhancement module, and the obtained frequency-domain information is better captured by the model than the original time-domain information in terms of periodicity, trend, and frequency features. Experiments on multiple benchmark datasets show that, compared with the state-of-the-art methods, the MSE and MAE of our model are improved by 11.1 and 6.36% on average, respectively, covering three applications (i.e. ETT, Exchange, and Weather). [ABSTRACT FROM AUTHOR]

Accurately forecasting the nonlinear degradation of lithium‐ion batteries (LIBs) using early‐cycle data can obviously shorten the battery test time, which accelerates battery optimization and production. In this work, a self‐adaptive long short‐term memory (SA‐LSTM) method has been proposed to predict the battery degradation trajectory and battery lifespan with only early cycling data. Specifically, two features were extracted from discharge voltage curves by a time‐series‐based approach and forecasted to further cycles using SA‐LSTM model. The as‐obtained features were correlated with the capacity to predict the capacity degradation trajectory by generalized multiple linear regression model. The proposed method achieved an average online prediction error of 6.00% and 6.74% for discharge capacity and end of life, respectively, when using the early‐cycle discharge information until 90% capacity retention. Furthermore, the importance of temperature control was highlighted by correlating the features with the average temperature in each cycle. This work develops a self‐adaptive data‐driven method to accurately predict the cycling life of LIBs, and unveils the underlying degradation mechanism and the importance of controlling environmental temperature. [ABSTRACT FROM AUTHOR]

Wind is one kind of high-efficient, environmentally-friendly, and cost-effective energy source. Wind power, as one of the largest renewable energy in the world, has been playing a more and more important role in supplying electricity. Though growing dramatically in recent years, the amount of generated wind power can be directly or latently affected by multiple uncertain factors, such as wind speed, wind direction, temperatures, and so on. More importantly, there exist very complicated dependencies of the generated power on the latent composition of these multiple time-evolving variables, which are always ignored by existing works and thus largely hinder the prediction performances. To this end, we propose DEWP, a novel Deep Expansion learning for Wind Power forecasting framework to carefully model the complicated dependencies with adequate expressiveness. DEWP starts with a stack-by-stack architecture, where each stack is composed of (i) a variable expansion block that makes use of convolutional layers to capture dependencies among multiple variables; (ii) a time expansion block that applies Fourier series and backcast/forecast mechanism to learn temporal dependencies in sequential patterns. These two tailored blocks expand raw inputs into different latent feature spaces which can model different levels of dependencies of time-evolving sequential data. Moreover, we propose an inference block corresponding for each stack, which applies multi-head self-attentions to acquire attentive features and maps expanded latent representations into generated wind power. In addition, to make DEWP more expressive in handling deep neural architectures, we adapt doubly residue learning to process stack-by-stack outputs. Accurate wind power forecasting (WPF) is then better achieved through fine-grained outputs by continuously removing stack residues and accumulating useful stack forecasts. Finally, we present extensive experiments in the real-world WPF application on two datasets from two different turbines, in order to demonstrate the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

In this study, we present the EEG-GCN, a novel hybrid model for the prediction of time series data, adept at addressing the inherent challenges posed by the data's complex, non-linear, and periodic nature, as well as the noise that frequently accompanies it. This model synergizes signal decomposition techniques with a graph convolutional neural network (GCN) for enhanced analytical precision. The EEG-GCN approaches time series data as a one-dimensional temporal signal, applying a dual-layered signal decomposition using both Ensemble Empirical Mode Decomposition (EEMD) and GRU. This two-pronged decomposition process effectively eliminates noise interference and distills the complex signal into more tractable sub-signals. These sub-signals facilitate a more straightforward feature analysis and learning process. To capitalize on the decomposed data, a graph convolutional neural network (GCN) is employed to discern the intricate feature interplay within the sub-signals and to map the interdependencies among the data points. The predictive model then synthesizes the weighted outputs of the GCN to yield the final forecast. A key component of our approach is the integration of a Gated Recurrent Unit (GRU) with EEMD within the GCN framework, referred to as EEMD-GRU-GCN. This combination leverages the strengths of GRU in capturing temporal dependencies and the EEMD's capability in handling non-stationary data, thereby enriching the feature set available for the GCN and enhancing the overall predictive accuracy and stability of the model. Empirical evaluations demonstrate that the EEG-GCN model achieves superior performance metrics. Compared to the baseline GCN model, EEG-GCN shows an average R2 improvement of 60% to 90%, outperforming the other methods. These results substantiate the advanced predictive capability of our proposed model, underscoring its potential for robust and accurate time series forecasting. [ABSTRACT FROM AUTHOR]

The building sector, known for its high energy consumption, needs to reduce its energy use due to rising greenhouse gas emissions. To attain this goal, a projection for domestic energy usage is needed. This work optimizes short-term load forecasting (STLF) in the building sector while considering several variables (energy consumption/generation, weather information, etc.) that impact energy use. It performs a comparative analysis of various machine learning (ML) models based on different data resolutions and time steps ahead (15 min, 30 min, and 1 h with 4-step-, 2-step-, and 1-step-ahead, respectively) to identify the most accurate prediction method. Performance assessment showed that models like histogram gradient-boosting regression (HGBR), light gradient-boosting machine regression (LGBMR), extra trees regression (ETR), ridge regression (RR), Bayesian ridge regression (BRR), and categorical boosting regression (CBR) outperformed others, each for a specific resolution. Model performance was reported using R 2 , root mean square error (RMSE), coefficient of variation of RMSE (CVRMSE), normalized RMSE (NRMSE), mean absolute error (MAE), and execution time. The best overall model performance indicated that the resampled 1 h 1-step-ahead prediction was more accurate than the 15 min 4-step-ahead and the 30 min 2-step-ahead predictions. Findings reveal that data preparation is vital for the accuracy of prediction models and should be model-adjusted. [ABSTRACT FROM AUTHOR]

The iron and steel industry significantly contributes to global energy use and greenhouse gas emissions. The rising deployment of volatile renewables and the resultant need for flexibility, coupled with specific challenges in electric steelmaking (e.g., operation optimization, optimized power purchasing, effective grid capacity monitoring), require accurate energy consumption and demand forecasts for electric steel mills to align with the energy transition. This study investigates diverse approaches to forecast the energy consumption and demand of an electric arc furnace—one of the largest consumers on the grid—considering various forecast horizons and objectives with limited knowledge on process parameters. The results are evaluated for accuracy, robustness, and costs. Two grid connection capacity monitoring approaches—a one-step and a multi-step Long Short-Term Memory neural network—are assessed for intra-hour energy demand forecasts. The one-step approach effectively models energy demand, while the multi-step approach encounters challenges in representing different operational phases of the furnace. By employing a combined statistic–stochastic model integrating a Seasonal Auto-Regressive Moving Average model and Markov chains, the study extends the forecast horizon for optimized day-ahead electricity procurement. However, the accuracy decreases as the forecast horizon lengthens. Nevertheless, the day-ahead forecast provides substantial benefits, including reduced energy balancing needs and potential cost savings. [ABSTRACT FROM AUTHOR]

Social networks and official statistics have become vital sources of information in times of health emergencies. The ability to monitor and profile social sentiment is essential for understanding public perception and response in the context of public health crises, such as the one resulting from the COVID-19 pandemic. This study will explore how social sentiment monitoring and profiling can be conducted using information from social networks and official statistics, and how this combination of data can offer a more complete picture of social dynamics in times of emergency, providing a valuable tool for understanding public perception and guiding a public health response. To this end, a three-layer architecture based on Big Data and Artificial Intelligence is presented: the first layer focuses mainly on collecting, storing, and governing the necessary data such as social media and official statistics; in the second layer, the representation models and machine learning necessary for knowledge generation are built, and in the third layer the previously generated knowledge is adapted for better understanding by crisis managers through visualization techniques among others. Based on this architecture, a KDD (Knowledge Discovery in Databases) framework is implemented using methodological tools such as sentiment analysis, fuzzy 2-tuple linguistic models and time series prediction with the Prophet model. As a practical demonstration of the proposed model, we use tweets as data source (from the social network X, formerly known as Twitter) generated during the COVID-19 pandemic lockdown period in Spain, which are processed to identify the overall sentiment using sentiment analysis techniques and fuzzy linguistic variables, and combined with official statistical indicators for prediction, visualizing the results through dashboards. [ABSTRACT FROM AUTHOR]

The existence of chaos is particularly relevant, as the identification of a chaotic behavior in a time series could lead to reliable short-term forecasting. This paper evaluates the existence of nonlinearity and chaos in the underlying process of the spot prices of the Spanish electricity market. To this end, we used daily data spanning from 1 January 2013, to 31 March 2021 and we applied a comprehensive framework that encompassed a wide range of techniques. Nonlinearity was analyzed using the BDS method, while the existence of a chaotic structure was studied through Lyapunov exponents, recurrence plots, and quantitative recurrence analysis. While nonlinearity was detected in the underlying process, conclusive evidence supporting chaos was not found. In addition, the generalized autoregressive conditional heteroscedastic (GARCH) model accounts for part of the nonlinear structure that is unveiled in the electricity market. These findings hold substantial value for electricity market forecasters, traders, producers, and market regulators. [ABSTRACT FROM AUTHOR]

Short-term electric load forecasting (STLF) plays a pivotal role in modern power system management, bolstering forecasting accuracy and efficiency. This enhancement assists power utilities in formulating robust operational strategies, consequently fostering economic and social advantages within the systems. Existing methods employed for STLF either exhibit poor forecasting performance or require longer computational time. To address these challenges, this paper introduces a hybrid learning approach comprising variational mode decomposition (VMD) and random vector functional link network (RVFL). The RVFL network, serving as a universal approximator, showcases remarkable accuracy and fast computation, owing to the randomly generated weights connecting input and hidden layers. Additionally, the direct links between hidden and output layers, combined with the availability of a closed-form solution for parameter computation, further contribute to its efficiency. The effectiveness of the proposed VMD-RVFL was assessed using electric load datasets obtained from the Australian Energy Market Operator (AEMO). Moreover, the effectiveness of the proposed method is demonstrated by comparing it with existing benchmark forecasting methods using two performance indices such as root mean square error (RMSE) and mean absolute percentage error (MAPE). As a result, our proposed method requires less computational time and yielded accurate and robust prediction performance when compared with existing methods. [ABSTRACT FROM AUTHOR]

Climate change can increase the frequency of extreme weather events and thus have a profound effect on the water flow, leading to occurrences such as floods and other natural disasters. Therefore, streamflow prediction under climate change is critical for regional risk management. This study introduces the Soil and Water Assessment Tool (SWAT), the Short and Long-term Memory (LSTM), the Gated Cycle Unit (GRU) and coupled Empirical Mode Decomposition (EMD) with LSTM/GRU, to investigate the feasibility of runoff forecast in the Dagu River basin, Jiaozhou Bay. The EMD technique breaks down the initial signal into multiple intrinsic modal functions that aid in capturing distinct data characteristics. These functions, when coupled with the robust mapping and learning capability of LSTM/GRU models, facilitate the prediction of runoff for time series data. In the prediction performance evaluation of the five models above, EMD-LSTM has the best performance with an R2 of 0.74 and RMSE 14.5 lower than the other models. Based on five GCMs in CMIP6, under three discharge scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5), the best-performed EMD-LSTM model is used to predict runoff changes of the study area under future climate change. The forecast results show significant increases in average annual runoff. Whether in mid or late of the century, the extreme streamflow will decrease from November to January and increase from February to May with the maximum amplitude of 41.96%, which implies the probability of spring floods. To some extents, this study proposes a method for improving the accuracy of runoff prediction and provides an early warning for possible regional flood disasters. [ABSTRACT FROM AUTHOR]

The extent of pollution in areas near industrial activities and Yamuna River has increased tremendously and to determine the overall water quality status in these locations is utmost important. This paper presents a comprehensive study on assessing the water quality and gain insights into the overall water quality status in the vicinity of industries and along the Yamuna River in Delhi. To achieve this, water samples were collected from Central Pollution Control Board (CPCB) for the last eight years (2013-2021) and was converted into a machine-readable format to facilitate further analysis. These samples were analysed for several water quality parameters, including pH, dissolved oxygen (DO), biochemical oxygen demand (BOD), chemical oxygen demand (COD), nutrient levels, heavy metals, and other relevant pollutants. By understanding these factors, a deeper understanding of the pollution sources and their impact on water quality is gained. Time series methods are applied to forecast future trends and values of the water quality parameters. This allows for predicting potential changes in water quality over time. A remarkable accuracy of 93.6% is attained in predicting water quality values up to the present time. The analysis reveals that water quality in the NCR region falls below acceptable standards. However, there has been a marginal enhancement in the water quality at Khajori Paltoon (Location L3) post-COVID. To achieve a more significant improvement in water quality, it is imperative to implement new policies. [ABSTRACT FROM AUTHOR]

In the dynamic landscape of cloud computing, efficient resource allocation and workload prediction are paramount for optimizing infrastructure utilization, cost management, and overall service quality. As a result, an efficient resource management approach involves pooling available resources among many customers in a manner that considers their energy usage, SLAs, and forecasting accuracy all at once. Based on the performance analysis results, it is determined that integrating these techniques would result in the most efficient and adaptable cloud data centre. We proposed a Model based on Deep Convolutional Neural Networks (DCNN) and Long Short-Term Memory (LSTM) for handling SLAs in the cloud from both the perspective of consumers and service providers. Subsequently, we delve into the methodology of applying deep CNN and LSTM models to the problem of workload prediction in cloud environments. This methodology encompasses data preprocessing, model architecture, training parameters, and the choice of performance metrics. To anticipate CPU consumption from time series data and detect SLA violations, we suggested a DCNN-LSTM model. The accuracy prediction, energy usage, CPU use, and Service Level Agreement monitoring are all a part of this model. The proposed method is effective in helping cloud providers cut down on service violations and associated fines. Regarding the composite metric of Energy SLA Violation, which assesses the combined aspects of energy use and adherence to Service Level Agreements (SLAs), DCNN-LSTM surpasses ARIMA-LSTM, CNN, LSTM, and ARIMA 6.8%, 10.88%, 16.6%, and 22.4%, respectively. [ABSTRACT FROM AUTHOR]

Predicting mortality rates is a crucial issue in life insurance pricing and demographic statistics. Traditional approaches, such as the Lee-Carter model and its variants, predict the trends of mortality rates using factor models, which explain the variations of mortality rates from the perspective of ages, gender, regions, and other factors. Recently, deep learning techniques have achieved great success in various tasks and shown strong potential for time-series forecasting. In this paper, we propose a modified Transformer architecture for predicting mortality rates in major countries around the world. Through the multi-head attention mechanism and positional encoding, the proposed Transformer model extracts key features effectively and thus achieves better performance in time-series forecasting. By using empirical data from the Human Mortality Database, we demonstrate that our Transformer model has higher prediction accuracy of mortality rates than the Lee-Carter model and other classic neural networks. Our model provides a powerful forecasting tool for insurance companies and policy makers. [ABSTRACT FROM AUTHOR]

As healthcare costs continue to pose significant challenges for governments and policymakers, accurate forecasting of medical inflation has become crucial in the European Union. This study aims to provide insights into the trajectory of medical inflation within the EU using the Autoregressive Integrated Moving Average (ARIMA) model and to check whether this model is an effective tool for predictions of medical inflation. The findings of the study have significant implications across various sectors. With accurate forecasts of medical inflation, policymakers can proactively address challenges, insurers can determine appropriate premiums and develop innovative models, and healthcare entities can allocate resources strategically to ensure financial stability and quality care. [ABSTRACT FROM AUTHOR]

Multivariate time series forecasting is essential in various fields, including healthcare and traffic management, but it is a challenging task due to the strong dynamics in both intra-channel relations (temporal patterns within individual variables) and inter-channel relations (the relationships between variables), which can evolve over time with abrupt changes. This paper proposes ERAN (Evolving Relational Attention Network), a framework for multivariate time series forecasting, that is capable to capture such dynamics of these relations. On the one hand, ERAN represents inter-channel relations with a graph which evolves over time, modeled using a recurrent neural network. On the other hand, ERAN represents the intra-channel relations using a temporal attentional convolution, which captures the local temporal dependencies adaptively with the input data. The elvoving graph structure and the temporal attentional convolution are intergrated in a unified model to capture both types of relations. The model is experimented on a large number of real-life datasets including traffic flows, energy consumption, and COVID-19 transmission data. The experimental results show a significant improvement over the state-of-the-art methods in multivariate time series forecasting particularly for non-stationary data. [ABSTRACT FROM AUTHOR]

Time series data prediction is an essential area of research in finance, and economics, among others. It involves analyzing and modeling data collected over time to make future predictions or forecast future trends. With the increasing availability of historical data and advancements in machine learning and deep learning techniques, time series data prediction has become an increasingly popular research topic in recent years. In this paper, we investigate the application of machine and deep learning models to time-series data for predicting the optimal time for trading stocks and options. Time-series data is defined as a collection of historical data points ordered by time, commonly used in predicting stock prices, stock indexes, and cryptocurrency prices. The study uses a publicly available dataset of Japanese stocks and options to train and test Support Vector Regression (SVR) and Long Short-Term Memory (LSTM) models. The goal of the research is to improve trading strategies by identifying the best times to buy and sell assets based on predictive models. The performance of the model are compared using three accuracy measurements: Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). The study has shown that The LSTM with dropout technique provided the best possible results with MSE 0.000124763, RMSE 0.011169727, and MAE 0.009058733. [ABSTRACT FROM AUTHOR]

In the research on strong alkali‐surfactant‐polymer flooding scaling prediction, the variation characteristics of ion concentration in the produced fluid are obviously consistent with those of scale component content. Consequently, studying the variation tendency of ion concentration in produced fluid is helpful for further revealing the scaling tendency. Given the periodicity and chaos characteristics of the ion concentration data of the produced fluid, an echo state network (ESN) is used to realize the relevant time series forecasting. Simultaneously, to cope with the failure of the ESN in selecting suitable reservoir parameters according to different characteristics of time series, a modified discrete particle swarm optimization algorithm based on objective space decomposition is used to optimize the reservoir parameters. The experimental results indicate that the improved ESN presents the lowest error and is the closest to the target value. [ABSTRACT FROM AUTHOR]

Copyright of Istanbul Commerce University Journal of Social Sciences / İstanbul Ticaret Üniversitesi Sosyal Bilimler Dergisi is the property of Istanbul Commerce University Journal of Social Sciences and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The advent of computer networks and the internet has drastically altered the means by which we share information and interact with each other. However, this technological advancement has also created opportunities for malevolent behavior, with individuals exploiting vulnerabilities to gain access to confidential data, obstruct activity, etc. To this end, intrusion detection systems (IDSs) are needed to filter malicious traffic and prevent common attacks. In the past, these systems relied on a fixed set of rules or comparisons with previous attacks. However, with the increased availability of computational power and data, machine learning has emerged as a promising solution for this task. While many systems now use this methodology in real-time for a reactive approach to mitigation, we explore the potential of configuring it as a proactive time series prediction. In this work, we delve into this possibility further. More specifically, we convert a classic IDS dataset to a time series format and use predictive models to forecast forthcoming malign packets. We propose a new architecture combining convolutional neural networks, long short-term memory networks, and attention. The findings indicate that our model performs strongly, exhibiting an F1 score and AUC that are within margins of 1% and 3%, respectively, when compared to conventional real-time detection. Also, our architecture achieves an ∼8% F1 score improvement compared to an LSTM (long short-term memory) model. [ABSTRACT FROM AUTHOR]

Local weather forecasts in the Arctic outside of settlements are challenging due to the dearth of ground-level observation stations and high computational costs. During winter, these forecasts are critical to help prepare for potentially hazardous weather conditions, while in spring, these forecasts may be used to determine flood risk during annual snow melt. To this end, a hybrid VMD-WT-InceptionTime model is proposed for multi-horizon multivariate forecasting of remote-region temperatures in Alaska over short-term horizons (the next seven days). First, the Spearman correlation coefficient is employed to analyze the relationship between each input variable and the forecast target temperature. The most output-correlated input sequences are decomposed using variational mode decomposition (VMD) and, ultimately, wavelet transform (WT) to extract time-frequency patterns intrinsic in the raw inputs. The resulting sequences are fed into a deep InceptionTime model for short-term forecasting. This hybrid technique has been developed and evaluated using 35+ years of data from three locations in Alaska. Different experiments and performance benchmarks are conducted using deep learning models (e.g., Time Series Transformers, LSTM, MiniRocket), and statistical and conventional machine learning baselines (e.g., GBDT, SVR, ARIMA). All forecasting performances are assessed using four metrics: the root mean squared error, the mean absolute percentage error, the coefficient of determination, and the mean directional accuracy. Superior forecasting performance is achieved consistently using the proposed hybrid technique. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Shenzhen University Science & Engineering is the property of Editorial Department of Journal of Shenzhen University Science & Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

We investigate the use of Generative Adversarial Networks (GANs) for probabilistic forecasting of financial time series. To this end, we introduce a novel economics-driven loss function for the generator. This newly designed loss function renders GANs more suitable for a classification task, and places them into a supervised learning setting, whilst producing full conditional probability distributions of price returns given previous historical values. Our approach moves beyond the point estimates traditionally employed in the forecasting literature, and allows for uncertainty estimates. Numerical experiments on equity data showcase the effectiveness of our proposed methodology, which achieves higher Sharpe Ratios compared to classical supervised learning models, such as LSTMs and ARIMA. [ABSTRACT FROM AUTHOR]

The confining pressure has a great effect on the internal force of the tunnel. During construction, the confining pressure which has a crucial impact on tunnel construction changes due to the variation of groundwater level and applied load. Therefore, the safety of tunnels must have the magnitude of confining pressure accurately estimated. In this study, a complete tunnel confining pressure time axis was obtained through high-frequency field monitoring, the data are segmented into a training set and a testing set. Using GRU and RNN models, a confining pressure prediction model was established, and the prediction results were analyzed. The results indicate that the GRU model has a fast-training speed and higher accuracy. On the other hand, the training speed of the RNN model is slow, with lower accuracy. The dynamic characteristics of soil pressure during tunnel construction require accurate prediction models to maintain the safety of the tunnel. The comparison between GRU and RNN models not only highlights the advantages of the GRU model but also emphasizes the necessity of balancing speed accuracy in tunnel construction confining pressure prediction modeling. This study is helpful in improving the understanding of soil pressure dynamics and developing effective prediction tools to promote safer and more reliable tunnel construction practices. [ABSTRACT FROM AUTHOR]

Recently, transfer learning has gained popularity in the machine learning community. Transfer Learning (TL) has emerged as a promising paradigm that leverages knowledge learned from one or more related domains to improve prediction accuracy in a target domain with limited data. However, for time series forecasting (TSF) applications, transfer learning is relatively new. This paper addresses the need for empirical studies as identified in recent reviews advocating the need for practical guidelines for Transfer Learning approaches and method designs for time series forecasting. The main contribution of this paper is the suggestion of a comprehensive framework for Transfer Learning Sensitivity Analysis (SA) for time series forecasting. We achieve this by identifying various parameters seen from various angles of transfer learning applied to time series, aiming to uncover factors and insights that influence the performance of transfer learning in time series forecasting. Undoubtedly, symmetry appears to be a core aspect in the consideration of these factors and insights. A further contribution is the introduction of four TL performance metrics encompassed in our framework. These TL performance metrics provide insight into the extent of the transferability between the source and the target domains. Analyzing whether the benefits of transferred knowledge are equally or unequally accessible and applicable across different domains or tasks speaks to the requirement of symmetry or asymmetry in transfer learning. Moreover, these TL performance metrics inform on the possibility of the occurrence of negative transfers and also provide insight into the possible vulnerability of the network to catastrophic forgetting. Finally, we discuss a sensitivity analysis of an Ensemble TL technique use case (with Multilayer Perceptron models) as a proof of concept to validate the suggested framework. While the results from the experiments offer empirical insights into various parameters that impact the transfer learning gain, they also raise the question of network dimensioning requirements when designing, specifically, a neural network for transfer learning. [ABSTRACT FROM AUTHOR]

This paper reviews the application of artificial neural network (ANN) models to time series prediction tasks. We begin by briefly introducing some basic concepts and terms related to time series analysis, and by outlining some of the most popular ANN architectures considered in the literature for time series forecasting purposes: feedforward neural networks, radial basis function networks, recurrent neural networks, and self-organizing maps. We analyze the strengths and weaknesses of these architectures in the context of time series modeling. We then summarize some recent time series ANN modeling applications found in the literature, focusing mainly on the previously outlined architectures. In our opinion, these summarized techniques constitute a representative sample of the research and development efforts made in this field. We aim to provide the general reader with a good perspective on how ANNs have been employed for time series modeling and forecasting tasks. Finally, we comment on possible new research directions in this area. [ABSTRACT FROM AUTHOR]

In recent years, researchers proposed a variety of deep learning models for wind power forecasting. These models predict the wind power generation of wind farms or entire regions more accurately than traditional machine learning algorithms or physical models. However, latest research has shown that deep learning models can often be manipulated by adversarial attacks. Since wind power forecasts are essential for the stability of modern power systems, it is important to protect them from this threat. In this work, we investigate the vulnerability of two different forecasting models to targeted, semi-targeted, and untargeted adversarial attacks. We consider a long short-term memory (LSTM) network for predicting the power generation of individual wind farms and a convolutional neural network (CNN) for forecasting the wind power generation throughout Germany. Moreover, we propose the Total Adversarial Robustness Score (TARS), an evaluation metric for quantifying the robustness of regression models to targeted and semi-targeted adversarial attacks. It assesses the impact of attacks on the model's performance, as well as the extent to which the attacker's goal was achieved, by assigning a score between 0 (very vulnerable) and 1 (very robust). In our experiments, the LSTM forecasting model was fairly robust and achieved a TARS value of over 0.78 for all adversarial attacks investigated. The CNN forecasting model only achieved TARS values below 0.10 when trained ordinarily, and was thus very vulnerable. Yet, its robustness could be significantly improved by adversarial training, which always resulted in a TARS above 0.46. [ABSTRACT FROM AUTHOR]

The time series analysis and prediction techniques are highly valued in many application fields, such as economy, medicine and biology, environmental sciences or meteorology, among others. In the last years, there is a growing interest in the sustainable and optimal management of a resource as scarce as essential: the water. Forecasting techniques for water management can be used for different time horizons from the planning of constructions that can respond to long-term needs, to the detection of anomalies in the operation of facilities or the optimization of the operation in the short and medium term. In this paper, a deep neural network is specifically designed to predict water consumption in the short-term. Results are reported using the time series of water consumption for a year and a half measured with 10-min frequency in the city of Murcia, the seventh largest city in Spain by number of inhabitants. The results are compared with K Nearest Neighbors, Random Forest, Extreme Gradient Boosting, Seasonal Autoregressive Integrated Moving Average and two persistence models as naive methods, showing the proposed deep learning model the most accurate results. [ABSTRACT FROM AUTHOR]

Forecasting exchange rates is a complex problem due to the inherent volatility and complex dynamics of exchange rates. Traditional forecasting models such as ARIMA often cannot capture these complexities especially for long-term forecasts. The objective of this study is to develop an accurate forecasting model for long-term exchange rates. A data set of eurodollar exchange rates from 2017 to 2022 was used for the present analysis. ARIMA and MLP models were developed and their performances were compared; the optimized MLP model equipped with 11 input neurons derived from significant lags achieved a scaled mean absolute error (MASE) of 0.75 on the test data while the MLP model significantly outperformed the ARIMA model, demonstrating its ability to capture underlying patterns and trends in the exchange rate data. The optimized MLP model also provided a 365-day forecast for 2023 exchange rates. The results of this study suggest that MLP models are a promising tool for long-term forecasting of exchange rates. Their ability to capture complex nonlinear relationships and adapt to changing market conditions makes them well suited for this challenging task. [ABSTRACT FROM AUTHOR]

IntroductionHypoxia is defined as a critically low-oxygen condition of water, which, if prolonged, can be harmful to fish and many other aquatic species. In the context of ocean salmon fish farming, early detection of hypoxia events is critical for farm managers to mitigate these events to reduce fish stress, however in complex natural systems accurate forecasting tools are limited. The goal of this research is to use a machine learning approach to forecast oxygen concentration and predict hypoxia events in marine net-pen salmon farms.MethodsThe developed model is based on gradient boosting and works in two stages. First, we apply auto-regression to build a forecasting model that predicts oxygen concentration levels within a cage. We take a global forecasting approach by building a model using the historical data provided by sensors at several marine fish farms located in eastern Canada. Then, the forecasts are transformed into binary probabilities that indicate the likelihood of a low-oxygen event. We leverage the cumulative distribution function to compute these probabilities.Results and discussionWe tested our model in a case study that included several cages across 14 fish farms. The experiments suggest that the model can detect future hypoxic events with a commercially acceptable false alarm rate. The resulting probabilistic predictions and oxygen concentration forecasts can help salmon farmers to prioritize resources, and reduce harm to crops.

Abstract Seasonal precipitation forecasts are essential for water resource management, agricultural activities, and the operational planning of hydropower systems. Any methodological advancement that enhances the accuracy of precipitation predictions will yield considerable societal benefits. In this context, this study proposes and evaluates two approaches for refining seasonal precipitation forecasts in Brazil, using simple data-based models, such as multiple linear regression (MLR) and nonlinear support vector machine (SVM). These models employ climate indices related to different teleconnection patterns that affect seasonal precipitation in Brazil, the unified gauge-based analysis of global daily precipitation from the Climate Prediction Center (CPC), and the precipitation forecasts from the Seasonal Forecast System 5 (SEAS5) as input variables. Both MLR and SVM models were validated from Jan-2017 to Dec-2020 using precipitation from the CPC as ground truth. The results suggest that, compared to SEAS5, MLR and SVM models enhance predictive accuracy and reduce bias in precipitation forecasts for the Southeast, Midwest, and North regions of Brazil during the austral summer. However, the performance of the models was found to be on par with the original predictions of SEAS5 in the Northeast and South regions, sectors of Brazil where the climate is significantly influenced by the El Niño-Southern Oscillation.

In recent years, the global installed capacity of wind power has grown rapidly. Wind power forecasting, as a key technology in wind turbine systems, has received widespread attention and extensive research. However, existing studies typically focus on the power prediction of individual devices. In the context of multi-turbine scenarios, employing individual models for each device may introduce challenges, encompassing data dilution and a substantial number of model parameters in power generation forecasting tasks. In this paper, a single-model method suitable for multi-device wind power forecasting is proposed. Firstly, this method allocates multi-dimensional random vectors to each device. Then, it utilizes space embedding techniques to iteratively evolve the random vectors into representative vectors corresponding to each device. Finally, the temporal features are concatenated with the corresponding representative vectors and inputted into the model, enabling the single model to accomplish multi-device wind power forecasting task based on device discrimination. Experimental results demonstrate that our method not only solves the data dilution issue and significantly reduces the number of model parameters but also maintains better predictive performance. Future research could focus on using more interpretable space embedding techniques to observe representation vectors of wind turbine equipment and further explore their semantic features.

This study brings about an innovative approach as rainfall forecast predominantly Boolean calculation is reviewed overall for 6 major cities of Pakistan for the time span of last quarter century. The study is aimed at improving the accuracy and the reliability of rainfall forecasting by making use of the capabilities of Artificial Intelligence (AI). This research investigates the efficacy of various machine learning models, including Naive Bayes, Logistic Regression, Support Vector Machines (SVM), K Nearest Neighbors (KNN), Gradient Boosting, and time series forecasting model ARIMA (Autoregressive Integrated Moving Average), for rainfall prediction. The model gets trained using 20 years of pakistan historical data performance where improving precision is the objective. With a highly scientific evaluation against real-world datasets, the new approach shows remarkable improvements in the accuracy of rainfall prediction compared to other conventional methods. Machine learning model KNN and time series forecasting model ARIM provide the good result in term of higher accuracy and lower RMSE. This results in the combination of environmental data science for innovation of the meteorological forecasting of this polarized area where the judgments are not so accurate.

Accurate predictive modeling of pandemics is essential for optimally distributing biomedical resources and setting policy. Dozens of case prediction models have been proposed but their accuracy over time and by model type remains unclear. In this study, we systematically analyze all US CDC COVID-19 forecasting models, by first categorizing them and then calculating their mean absolute percent error, both wave-wise and on the complete timeline. We compare their estimates to government-reported case numbers, one another, as well as two baseline models wherein case counts remain static or follow a simple linear trend. The comparison reveals that around two-thirds of models fail to outperform a simple static case baseline and one-third fail to outperform a simple linear trend forecast. A wave-by-wave comparison of models revealed that no overall modeling approach was superior to others, including ensemble models and errors in modeling have increased over time during the pandemic. This study raises concerns about hosting these models on official public platforms of health organizations including the US CDC which risks giving them an official imprimatur and when utilized to formulate policy. By offering a universal evaluation method for pandemic forecasting models, we expect this study to serve as the starting point for the development of more accurate models.

Time series forecasting is an essential tool across numerous domains, yet traditional models often falter when faced with unilateral boundary conditions, where data is systematically overestimated or underestimated. This paper introduces a novel approach to the task of unilateral boundary time series forecasting. Our research bridges the gap in existing methods by proposing a specialized framework to accurately forecast within these skewed datasets. The cornerstone of our approach is the unilateral mean square error (UMSE), an asymmetric loss function that strategically addresses underestimation biases in training data, improving the precision of forecasts. We further enhance model performance through the implementation of a dual model structure that processes underestimated and accurately estimated data points separately, allowing for a nuanced analysis of the data trends. Additionally, feature reconstruction is employed to recapture obscured dynamics, ensuring a comprehensive understanding of the data. We demonstrate the effectiveness of our methods through extensive experimentation with LightGBM and GRU models across diverse datasets, showcasing superior accuracy and robustness in comparison to traditional models and existing methods. Our findings not only validate the efficacy of our approach but also reveal its model-independence and broad applicability. This work lays the groundwork for future research in this domain, opening new avenues for sophisticated analytical models in various industries where precise time series forecasting is crucial.

Automated Machine Learning (AutoML) tools for time series forecasting represent a frontier in both academic and industrial research, addressing the need for efficient, accurate predictions in various domains. This study focuses on the development of Automated Time Series Forecasting (AutoTS), specifically in the domains of demand forecasting and traffic prediction. Through a comprehensive empirical evaluation and comparative analysis, this study investigates the performance of 35 time series forecasting methods and scrutinizes the forecast quality and time series characteristics. In the experiments, we used six real and one synthetic datasets with 325 time series in total. The forecasting for four distinct horizons is studied. From our large-scale empirical study, we draw the following main conclusions: boosting-based methods, which are often overlooked, have a strong performance; the global modeling approach is promising because it provides competitive performance with a small computational cost; meta-learning via portfolio selection performs better than one based on meta-features. We hope that our empirical study will pave the way for more efficient AutoML systems for time series forecasting.

Video prediction presents a formidable challenge, requiring effectively processing spatial and temporal information embedded in videos. While recurrent neural network (RNN) and transformer-based models have been extensively explored to address spatial changes over time, recent advancements in convolutional neural networks (CNNs) have yielded high-performance video prediction models. CNN-based models offer advantages over RNN and transformer-based models due to their ease of parallel processing and lower computational complexity, highlighting their significance in practical applications. However, existing CNN-based video prediction models typically treat the spatiotemporal channels of videos similarly to the channel axis of static images. They stack frames in temporal order to construct a spatiotemporal axis and employ standard $1\times 1$ convolution operations. Nevertheless, this approach has its limitations. Applying $1\times 1$ convolution directly to the spatiotemporal axis results in a mixture of temporal and spatial information, which may lead to computational inefficiencies and reduced accuracy. Additionally, this operation needs to improve in processing temporal data. This study introduces a CNN-based time series decomposition model for video prediction. The proposed model first divides the $1\times 1$ convolution operation within the channel aggregation module to independently process the temporal and spatial dimensions. To capture evolving features, the temporal axis is segregated into trend and residual components, followed by applying a time series decomposition forecasting method. To assess the performance of the proposed technique, experiments were conducted using the moving MNIST, KTH, and KITTI-Caltech benchmark datasets. In the experiments on moving MNIST, despite a reduction of approximately 55% in the number of parameters and 37% in computational cost, the proposed method improved accuracy by up to 7% compared to the previous approach.

Myoelectric indices forecasting is important for muscle fatigue monitoring in wearable technologies, adaptive control of assistive devices like exoskeletons and prostheses, functional electrical stimulation (FES)-based Neuroprostheses, and more. Non-stationary temporal development of these indices in dynamic contractions makes forecasting difficult. This study aims at incorporating transfer learning into a deep learning model, Myoelectric Fatigue Forecasting Network (MEFFNet), to forecast myoelectric indices of fatigue (both time and frequency domain) obtained during voluntary and FES-induced dynamic contractions in healthy and post-stroke subjects respectively. Different state-of-the-art deep learning models along with the novel MEFFNet architecture were tested on myoelectric indices of fatigue obtained during ${a}\text {)}$ voluntary elbow flexion and extension with four different weights (1 kg, 2 kg, 3 kg, and 4 kg) in sixteen healthy subjects, and ${b}\text {)}$ FES-induced elbow flexion in sixteen healthy and seventeen post-stroke subjects under three different stimulation patterns (customized rectangular, trapezoidal, and muscle synergy-based). A version of MEFFNet, named as pretrained MEFFNet, was trained on a dataset of sixty thousand synthetic time series to transfer its learning on real time series of myoelectric indices of fatigue. The pretrained MEFFNet could forecast up to 22.62 seconds, 60 timesteps, in future with a mean absolute percentage error of 15.99 ± 6.48% in voluntary and 11.93 ± 4.77% in FES-induced contractions, outperforming the MEFFNet and other models under consideration. The results suggest combining the proposed model with wearable technology, prosthetics, robotics, stimulation devices, etc. to improve performance. Transfer learning in time series forecasting has potential to improve wearable sensor predictions.

Time series forecasting involves justifying assertions scientifically regarding potential states or predicting future trends of an event based on historical data recorded at various time intervals. The field of time series forecasting, supported by diverse deep learning models, has made significant advancements, rendering it a prominent research area. The broad spectra of available time series datasets serve as valuable resources for conducting extensive studies in time series analysis with varied objectives. However, the complexity and scale of time series data present challenges in constructing reliable prediction models. In this paper, our objectives are to introduce and review methodologies for modeling time series data, outline the commonly used time series forecasting datasets and different evaluation metrics. We delve into the essential architectures for trending an input dataset and offer a comprehensive assessment of the recently developed deep learning prediction models. In general, different models likely serve different design goals. We boldly examine the performance of these models under the same time series input dataset with an identical hardware computing system. The measured performance may reflect the design flexibility among all the ranked models. And through our experiments, the SCINet model performs the best in accuracy with the ETT energy input dataset. The results we obtain could give a glimpse in understanding the model design and performance relationship. Upon concluding the paper, we shall provide further discussion on future deep learning research directions in the realm of time series forecasting.

Forecasting crowd congestion is crucial for ensuring comfortable mobility and public safety. Existing methods forecast crowding by capturing the increase in planned visits, which facilitates the methods in estimating the start of crowding. However, forecasting the change in the degree of crowding until the end is challenging owing to the lack of visitors’ return plans and the deviation of visitor movements from preannounced event schedules. To address this issue, this study developed a novel framework for forecasting the start of crowding and its change over time (termed the lifespan of crowded events (LCE)). Based on the concept that event purposes influence the crowding patterns, our framework models these patterns according to the event purposes. Inspired by the acoustic synthesis that can successfully model the change in the sound volume for each instrument, we extended a canonical long short-term memory (LSTM) model with the concept of ADSR envelope, wherein the sound (crowd) volume changes can be represented within simple state transitions. The proposed versatile acoustic tri-state envelope for segmental LSTM, namely VATES, is evaluated on two datasets: synthetic and real-world mobility datasets. The results demonstrate that VATES can forecast crowding patterns with a 24.3% performance improvement, and precisely predict the start and end times of crowding, thereby improving by 6.6% and 26.1% respectively. We believe that our method enhances urban safety and mobility in crowded events, contributing to smarter city management.

Realistic anomalous data noise is usually locally distributed in nature. To address the challenge posed by clustered anomalous noise in time series data, which not only reduces the accuracy of forecasting models but also presents a unique challenge due to its clustered nature and existing methods’ limitations in handling such noise, this study introduces a novel denoising algorithm leveraging an optimized Long Short-Term Memory (LSTM) network, named ‘Denoising Time Window Optimized LSTM’ (DTW-LSTM), which innovatively identifies and rectifies clustered anomalous noise in time series data, significantly enhancing prediction accuracy by adeptly balancing the denoising process and data integrity preservation, as evidenced by comprehensive experimental validations. The algorithm employs a sliding time window mechanism to analyze the original series and its Fourier approximation, identifying clustered anomalies through a clustering model. It then uses a symmetrical data mixing and crossover technique to enhance the data quality in noisy segments. This process preserves the integrity of time series data by maintaining a balance between the original and fitted series, leading to a significant improvement in prediction accuracy. Experimental results indicate that the algorithm can adeptly discern the presence of clustered anomalous noise within the current time window and effectively rectify sections of data containing such noise. By surpassing the performance benchmarks set by contemporary forecasting models, this algorithm has established itself as the new SOTA method in the field of time series forecasting. Each sub-model contributes to the superior accuracy and reliability of the algorithm.

In recent years, time series forecasting has been widely used in various fields, especially in financial markets. Stock trend forecasting has become one of the most common and complex challenges faced by investors and researchers. However, much of the current research relies primarily on single-granularity stock data for forecasting, with relatively few studies on multi-granularity data and fewer studies on spatial correlation of multi-granularity data. This inherent limitation restricts the comprehensive extraction of valuable information. To address this challenge, we propose the Multi-Granularity Deep Spatio-Temporal Correlation Framework (MDSTCF). Our approach combines the strengths of a multi-granularity residual learning, gated recurrent units, and graph attention networks to extract spatio-temporal information specific to each granularity. Subsequently, predictions at each granularity are generated through the prediction layer. Finally, a soft attention mechanism is employed to assign weights to the predictions at each granularity to obtain the final result. Comprehensive experiments conducted on tow stock datasets show that the proposed forecasting model improves the F1 score by about 7.88% and 11.2%, and the cumulative relative returns are close to 80% and 40%, respectively, compared to the previously studied time series forecasting models. The results clearly indicates that fusing multi-granularity information can significantly improve the performance of time series forecasting.

Local weather forecasts in the Arctic outside of settlements are challenging due to the dearth of ground-level observation stations and high computational costs. During winter, these forecasts are critical to help prepare for potentially hazardous weather conditions, while in spring, these forecasts may be used to determine flood risk during annual snow melt. To this end, a hybrid VMD-WT-InceptionTime model is proposed for multi-horizon multivariate forecasting of remote-region temperatures in Alaska over short-term horizons (the next seven days). First, the Spearman correlation coefficient is employed to analyze the relationship between each input variable and the forecast target temperature. The most output-correlated input sequences are decomposed using variational mode decomposition (VMD) and, ultimately, wavelet transform (WT) to extract time-frequency patterns intrinsic in the raw inputs. The resulting sequences are fed into a deep InceptionTime model for short-term forecasting. This hybrid technique has been developed and evaluated using 35+ years of data from three locations in Alaska. Different experiments and performance benchmarks are conducted using deep learning models (e.g., Time Series Transformers, LSTM, MiniRocket), and statistical and conventional machine learning baselines (e.g., GBDT, SVR, ARIMA). All forecasting performances are assessed using four metrics: the root mean squared error, the mean absolute percentage error, the coefficient of determination, and the mean directional accuracy. Superior forecasting performance is achieved consistently using the proposed hybrid technique.