Search

Autonomous vehicles (AV)s have become a trending topic nowadays since they have the potential to solve traffic problems, such as accidents and congestion. Although AV systems have greatly evolved, it still have their limitations. For example, Google reported that their AVs have been involved in several collisions and near misses. While most of these collisions and near misses were caused by third parties, the AVs should be able to predict and avoid them. Events like this show that there is still room for improvement in the AV system. This paper aims to present a review of the state-of-the-art algorithms proposed to enable AV behaviour prediction systems to predict trajectories and intentions for pedestrians and vehicles. This will be achieved by using information from previous literature review papers, recent works, and results obtained using well-known datasets. • An AV behaviour prediction system must have a long (> 3s) prediction horizon. • It is a challenging task to model interaction between agents. • Datasets from onboard sensors should be used to predict road agents' behaviour. • Deep learning techniques have shown better results to predict behaviours. • Algorithms should be evaluated using the same datasets and evaluation metrics. [ABSTRACT FROM AUTHOR]

Grey prediction evolutionary algorithm (GPE) is an emerging category of meta-heuristic algorithm based on mathematical model. It generates offspring by predicting the macroscopic evolutionary trend of population series. On the contrary, the mining of local evolutionary trend are flawed. To address this imperfection, this paper introduces a quadratic function as a surrogate model to mine the local evolutionary trend of population series for the first time. Thereby an improved grey prediction evolutionary algorithm called grey prediction evolutionary algorithm with a surrogate model based on quadratic interpolation (GPE-QI) is designed. The surrogate model based on quadratic interpolation is regarded as an external auxiliary tool to mine the local evolutionary trend of population series by using currently superior individuals. Through utilizing the mined local trend information, a donor population closer to the optimal solution is formed to improve the prediction direction of the algorithm. The performance of GPE-QI is tested on CEC2019, CEC2020 benchmark functions and six engineering design problems. The experimental results demonstrate the stronger competitiveness of GPE-QI to well-established algorithms in terms of solving accuracy and convergence rate. If this paper is accepted, MATLAB codes associated with this paper will be uploaded to https://github.com/Zhongbo-Hu/Prediction-Evolutionary-Algorithm-HOMEPAGE. • Improvement can be made by enhancing local search of grey prediction evolution. • Mine local evolutionary trend of population series by quadratic interpolation. • Construct a surrogate model by using the local evolutionary trend. • Design an improved grey prediction evolutionary algorithm with a surrogate model. • The proposed algorithm is more competitive than the well-established. [ABSTRACT FROM AUTHOR]

The accurate prediction of digging load serves as a fundamental cornerstone for advancing the development of intelligent and unmanned excavators. Given the complex nonlinear dynamics of digging load, this paper proposes a novel prediction model for excavator load power based on the chaos theory and support vector regression (SVR). The presence of chaos in the dynamic digging load system is detected through phase space reconstruction. SVR is utilized for nonparametric modeling and prediction, with the reconstructed phase space capturing the essential characteristics of excavator load and serving as inputs for SVR. To optimize the hyperparameters, an improved particle swarm optimization (IPSO) algorithm is presented. Excavation experiments conducted under two typical load conditions demonstrate the superiority of the proposed chaos-based IPSO-SVR model in terms of prediction accuracy. This research lays a solid foundation for practical load prediction in industrial excavator settings. [ABSTRACT FROM AUTHOR]

• We propose a novel hybrid model namely MEEMD-Mix-KELM to forecast crude oil price. • The emerging MEEMD, with mode-splitting prevention, is used to decompose oil price. • We create a new mix-kernel for ELM by absorbing merits of local and global kernels. • Empirical results show that the proposed hybrid model achieves significant effects. It is of vital importance for governments, enterprises, and investors to forecast crude oil prices accurately, while this task is beset with difficulties and challenges due to the complex patterns in oil prices. This paper aims to propose a novel hybrid method to model and forecast the crude oil price by integrating median ensemble empirical mode decomposition (MEEMD) and mix-kernel extreme learning machine (Mix-KELM). Firstly, the emerging MEEMD is employed to decompose the crude oil price into several simple subseries. Secondly, a novel mix-kernel is developed for extreme learning machine (ELM) by combining the advantage of the local kernel (i.e., Radial Basis Function in learning ability) and global kernel (i.e., Sigmoid in generalization ability), with weights of the kernels optimized through genetic algorithm. Thirdly, the proposed Mix-KELM is applied to forecast the subseries of crude oil price, and the sub-forecasting results are integrated to generate the final results. The empirical results show that our proposed MEEMD-Mix-KELM model with different forecasting horizons significantly outperforms the benchmarks in terms of forecasting accuracy and robustness test. Taking one-step-ahead forecasting as an example, the proposed model exhibits the lowest prediction errors in terms of mean absolute error, symmetric mean absolute percentage error, and root mean squared error with values of 1.1767, 0.0135, and 1.5717, respectively. [ABSTRACT FROM AUTHOR]

The uncertainty associated with option price predictions has largely been overlooked in the literature. This paper aims to fill this gap by quantifying such uncertainty using conformal prediction. Conformal prediction is a model-agnostic procedure that constructs prediction intervals, ensuring valid coverage in finite samples without relying on distributional assumptions. Through the simulation of synthetic option prices, we find that conformal prediction generates prediction intervals for gradient boosting machines with an empirical coverage close to the nominal level. Conversely, non-conformal prediction intervals exhibit empirical coverage levels that fall short of the nominal target. In other words, they fail to contain the actual option price more frequently than expected for a given coverage level. As anticipated, we also observe a decrease in the width of prediction intervals as the size of the training data increases. However, we uncover significant variations in the width of these intervals across different options. Specifically, out-of-the-money options and those with a short time-to-maturity exhibit relatively wider prediction intervals. Then, we perform an empirical study using American call and put options on individual stocks. We find that the empirical results replicate those obtained in the simulation experiment. • Conformal prediction quantifies well the uncertainty of option price predictions. • Conformal prediction intervals have empirical coverage near the nominal level. • Non-conformal prediction intervals have empirical coverage below the nominal level. • Large variations in prediction intervals are found for American call and put options. • Larger intervals are found for out-of-the-money and short time-to-maturity options. [ABSTRACT FROM AUTHOR]

The ongoing transition in the electricity market, coupled with the rising penetration of renewable energy sources, has spurred demand for enhanced load forecasting accuracy. Traditional single predictors exhibit limited robustness and diversity, rendering them unsuitable for most forecasting situations. Moreover, weather conditions, holidays, and social events significantly influence load demand, making forecasting techniques that account for these factors increasingly relevant. To bridge the gaps and further enhance predictive performance, in this paper, a novel adaptive multivariate combined forecasting system is proposed, incorporating redundant feature elimination, adaptive model selection, multi-objective intelligent optimization with non-negative constraints, and reliable interval prediction theory. The proposed system employs non-linear feature selection to identify effective load features and develops an advanced Multi-Objective Optimization Strategy (MOS) for combining multiple optimal predictors under various scenarios. The weight coefficients of sub-predictors are theoretically demonstrated to be Pareto optimal, and the MOS exhibits superior distribution and convergence. Simulation results reveal that the developed system substantially outperforms benchmark models, achieving reliable uncertainty quantification, which facilitates the rational allocation of spinning reserves and enhances dispatching efficiency. [ABSTRACT FROM AUTHOR]

Federated Learning (FL) presents a decentralized approach to model training in the agri-food sector and offers the potential for improved machine learning performance, while ensuring the safety and privacy of individual farms or data silos. However, the conventional FL approach has two major limitations. First, the heterogeneous data on individual silos can cause the global model to perform well for some clients but not all, as the update direction on some clients may hinder others after they are aggregated. Second, it is lacking with respect to the efficiency perspective concerning communication costs during FL and large model sizes. This paper proposes a new technical solution that utilizes network pruning on client models and aggregates the pruned models. This method enables local models to be tailored to their respective data distribution and mitigate the data heterogeneity present in agri-food data. Moreover, it allows for more compact models that consume less data during transmission. We experiment with a soybean yield forecasting dataset and find that this approach can improve inference performance by 15.5% to 20% compared to FedAvg, while reducing local model sizes by up to 84% and the data volume communicated between the clients and the server by 57.1% to 64.7%. Our method demonstrates the potential to use efficient models that are more environmentally friendly to support the agri-food sector's transition to net zero. Future enhancements of this method could further optimize distributed learning in agri-food, enhancing sustainability and applicability. • We propose a new federated pruning learning method for yield forecasting. • Our method can be used to improve local inference performance. • The method reduces communication costs during training and model sizes. • Our method facilitates the on-edge implementation of ML models using sparse tensor. • The method is generalizable and works with different pruning policies and schedules. [ABSTRACT FROM AUTHOR]

This paper proposes a new model for creating fuzzy time series and conducting future forecasts using the Bi-directional Long Short-term Memory model (Bi-LSTM). This study contributes by using both classical machine learning and deep learning techniques. Firstly, a self-updating clustering algorithm is used to determine the optimal number of clusters for time series. Secondly, the study describes a method for determining the fuzzy relationship between elements of time series and the identified clusters. Thirdly, the developed model creates the fuzzy time series based on the established rules. Finally, forecasting for the future value using the Bi-LSTM model with some improvement in optimizing the parameters and increasing the accuracy of forecasting result. The steps of the proposed model are clearly illustrated by a numerical example. Moreover, this model also tested on various datasets, including M3, M4, and M5 datasets. By evaluating its performance using parameters such as sMAPE and RMSSE, the proposed model has shown significant enhancements when compared to existing models. Additionally, the model's effectiveness is further demonstrated through its successful application in forecasting the VN index stock in Vietnam. [ABSTRACT FROM AUTHOR]

The focus of this paper is to improve short-term load forecasting for electric power. To achieve this goal, the study explores and evaluates hybrid models, specifically using the CatBoost and XGBoost algorithms, which are optimized with different optimizers. The study incorporates hourly electricity load data and also includes temperature data to enhance the precision of the forecasting models. Statistical metrics are then used to assess the performance of these models. The study evaluates the performance of the hybrid models on both training and testing datasets. It finds that the CatBoost-Arithmetic Optimization Algorithm hybrid model outperforms the other models in the training dataset. However, in the testing dataset, the XGBoost- Arithmetic Optimization Algorithm hybrid model demonstrates superior performance compared to the CatBoost models. The study conducts an importance and sensitivity analysis to understand which variables have the most significant impact on the target variable, which is likely electricity load. The results of this analysis reveal that temperature is the most influential variable affecting the target variable. Additionally, the month variable is identified as having a notable impact on load forecasting. These findings suggest that employing hybrid models, particularly those optimized with appropriate algorithms, can significantly improve the accuracy of short-term load forecasting. Moreover, the study highlights the importance of incorporating temperature data into these models, as temperature is a key driver of electricity load patterns. [ABSTRACT FROM AUTHOR]

Telemonitoring of Parkinson's disease has important implications for early diagnosis and treatment of patients. Most of the existing feature selection methods for remote prediction of PD severity are based on correlation and rarely consider causality, thus compromising the robustness of the model. Therefore, a causal game-based feature selection (CGFS) model is proposed for remote PD symptom severity assessment. Firstly, to address the challenge of small data size, the similar patient transfer strategy is designed to find data from source domain patients with conditions similar to those of the target patient. Secondly, the undirected equivalent greedy search method is employed to construct the causal graph between features and PD severity scores, and the robustness of the model is improved by selecting causal features. Then, to enhance the prediction accuracy, this paper utilizes the cooperative game approach Shapley value to evaluate the contribution of neighborhood nodes of the target value, and selects the features with causality and high contribution to form the final feature subset. Finally, the subset is input into the random forest to further enhance robustness and performance of the model. Experiments on Parkinson's telemonitoring dataset and the tapping dataset with different biomarkers show that the robustness of the feature subset selected by the CGFS model, and the prediction performance is better than advanced models compared. Therefore, the validity and universality of the CGFS method is demonstrated in remote PD severity prediction. [ABSTRACT FROM AUTHOR]

Enhancing the accuracy of breast cancer recurrence prediction is crucial, mainly when dealing with genomics data, which presents challenges such as high dimensionality, noise, non-linearity, and limited sample sizes. This paper introduces Semi-Supervised Survival Laplacian Regression (S3LR), a novel feature selection algorithm designed to improve breast cancer recurrence prediction by effectively handling censored, event, and unlabeled data. S3LR modifies the Laplacian Score (LS) by incorporating a distance matrix to calculate the weight matrix, and it integrates heuristic and metaheuristic optimization algorithms to optimize the weighted matrix. These enhancements refine feature selection and overall performance. In our evaluations using three datasets and comparisons with state-of-the-art techniques, S3LR combined with Particle Swarm Optimization (PSO) demonstrates significant improvements in C-index and mean absolute error (MAE). Average C-index values reach 68.80 %, 59.49 %, and 67.66 %, with average MAE values of 15.98, 7.87, and 8.65 months, respectively. These results showcase S3LR's effectiveness in predicting recurrence, even with censored data, for more precise and reliable outcomes. Furthermore, the framework's versatility extends beyond breast cancer and can readily be applied to address other survival and recurrence problems. [ABSTRACT FROM AUTHOR]

In the context of today's energy shortage, wind energy plays a crucial role as one of the most widely used renewable energy sources. However, in order to fully utilize the potential of wind energy, it is necessary to build a precise and reliable wind-speed prediction system. Conventional wind-speed prediction models frequently rely on a solitary model for forecasting, and sometimes these models cannot accurately fit the nonlinear characteristics of the data, resulting in poor prediction performance. Therefore, to address this deficiency, a new wind-speed point-interval prediction system is proposed in this study. The fuzzy information granulation technology is adopted in this system to reduce the dimension of data and solve the problem of redundancy in wind-speed data. In addition, this paper uses the multi-objective dragonfly algorithm and combines sub-models using the multi-nonlinear weight strategy to address the issue of inadequate precision exhibited by an individual prediction model. In order to verify the effectiveness of this wind-speed prediction system, the Penglai wind farm in Shandong Province is used as the dataset to construct the model, and several reference models are selected to conduct point prediction and interval prediction comparison experiments respectively. The experimental results show that the combined prediction system has the best prediction performance, with MAPE values of 14.16%, 15.02% and 15.74%, respectively. Compared with other benchmark models, the average improvement rates for the multi-step prediction were 14.97%, 16.48% and 19.11%, respectively. [ABSTRACT FROM AUTHOR]

Emergency Department overcrowding is a well-known problem. The consequences are long waiting times for patients, reduced service quality, and the potential for increased mortality rates. Moreover, this problem is escalating with the aging of the population and the chronification of diseases. Being aware of such concerns, Emergency Departments started to conduct patient accountability years ago, and now they are ready to take advantage of the gathered data to improve the quality of service to citizens. However, to achieve this goal, they need the technology that enables the generation of predictive models from such data to improve the planning of their resources. In this paper, we face the task of forecasting patient admissions by using Deep Neural Networks, particularly a modern Attention-based model. As well, historical records do not provide a full picture of how patients attend the emergency service. Relying on the possibilities given by the attention mechanism, we also propose the use of exogenous information such as calendar data, weather, air quality, allergens, and information extracted from the web via Google Trends to enhance admissions prediction accuracy. In this regard, we have tested different configurations of the exogenous variables to isolate which ones provide relevant information that improves the model. In our experiments, we have seen that the calendar, weather, and air quality provide the most valuable information, meanwhile, allergens and Google Trends data appear to be hindering the models' performance. • Using Attention model for forecasting Emergency Department (ED) patient admissions. • Comparing Attention model with LSTM-based networks for ED patients admissions. • Exogenous variables: weather, calendar, air quality, allergens and Google Trends. • Analyzing the impact of each type of exogenous variable in the forecast. • Attention model with calendar data achieves the best results. [ABSTRACT FROM AUTHOR]

Accurate photovoltaic (PV) power forecast is crucial for carbon neutrality. Current researches on PV power forecast mainly focus on using temporal information from single PV station, and the spatial information in multiple PV power stations are often neglected. To address this problem, this paper introduces Moran index to verify the spatial autocorrelation of PV power for the first time, uses Granger causality test and transfer entropy to reveal the spatial information gain in PV power forecast for the first time, and proposes a novel spatial–temporal probabilistic PV forecast method using deep Graph Gated Recurrent Unit (GraphGRU) network-based spatial–temporal multi-task learning and Kernel Density Estimation (KDE). Deep GraphGRU combines the advantages of Graph Convolutional Network (GCN) in spatial feature extraction and the advantages of Gated Recurrent Unit (GRU) network in temporal feature extraction, and thus has strong ability to extract spatial–temporal information in historical data of multiple different PV power stations. Through GraphGRU, temporal dependency information extracted from historical data of multiple PV stations can promote each other to improve the forecast accuracy of each PV stations. KDE is used for estimating the joint probabilistic density function and giving the spatial–temporal probabilistic confidence interval of PV power. Experiments were performed in the five-year actual PV power data from 11 provinces of Belgium and the three-year solar irradiation data from 12 places in China to verify the superiorities of the proposed method. Comparison with conventional spatial–temporal and temporal forecast methods show that the proposed GraphGRU-based spatial–temporal forecast method can extract the spatial–temporal information from multiple PV power stations well and significantly outperform conventional temporal forecast methods. [ABSTRACT FROM AUTHOR]

Patient outcomes forecasting (POF) has been shown to be an effective diagnostic assistant that can be used to predict disease progression and patient status in advance. In practice, the variation in patient progress and physical symptoms poses significant challenges for POF model training and testing. These differences result in POF models trained on one set of patients (source domain) not being efficiently self-adapted to the disease prediction task of another set of patients (target domain). Domain adaptive approaches can balance the overall distribution differences, named long-term divergence between patient training data and test data. However, the complex time series dependence and multivariate distribution properties of patient clinical data lead to the inability of existing domain adaptation methods to guarantee consistent distribution of patient clinical data at a finer granularity, named short-term divergence, e.g., clinical data distribution at different time intervals. In this paper, we propose a novel POF approach (TADA) based on clinical time series data using T emporal-aware A dversarial D omain A daptation. In TADA, the traditional binary domain labels are generalized to act as the supervision signals to adversarially bridge the domain gap considering both long-term and short-term clinical information. TADA not only preserves the long-term representation of the clinical data but also captures the short-term data distribution at each period. The experimental results, based on both the publicly available and private clinical time series datasets, demonstrate that TADA can provide more reliable forecasting than state-of-the-art approaches. • Both long-term and short-term domain gap are simultaneously addressed. • A temporal-aware adversarial adaptation model to handle clinical time series data. • Theoretical analysis is provided for the patient outcomes forecasting model. [ABSTRACT FROM AUTHOR]

Accurately predicting support resistance in coal mine working faces is a pressing matter in the field of intelligent mining research, which holds immense importance in enhancing the safety of coal production and the overall level of intelligent mining. The current research in this field primarily focuses on utilizing deep temporal sequence models for prediction. While there have been promising advancements in this area, concerns persist regarding the dependability of predicted outcomes. In fact, the movement of overburden in mining faces varies over time and space, and the changes in the support resistance that reflect the instability of overburden movement also have temporal and spatial characteristics. Therefore, in order to improve the reliability of model prediction, this paper proposes the application of deep spatiotemporal sequence models for support resistance prediction, and identifies the PredRNN and PredRNN++ models as suitable options. The exploration on the application of deep spatiotemporal sequence models for predicting the support resistance has been conducted. The proposed application involves the establishment of a fully working face area grid model with support number and mining cycle as coordinate units. Different from deep temporal sequence models that take the studied grid features as observations, the 25 adjacent spatiotemporal feature grids corresponding to the studied grid are considered as observations for the deep spatiotemporal sequence models. Based on the predicted three-dimensional tensor, the required predicted support resistance data can be obtained. Based on the complete support resistance data from 9 working faces in a coal mine in Datong mining area from 2009 to 2018, a comprehensive dataset was established to experimentally analyze the predictive performance and reliability of the LSTM, PredRNN, and PredRNN++ models. The experimental results demonstrate significant improvements in predictive performance and reliability of deep spatiotemporal sequence models compared to the current mainstream deep temporal sequence models. In terms of predictive performance, the PredRNN and PredRNN++ models have improved their feature extraction ability by incorporating the convolution operations to extract the spatial correlation of the input data. Furthermore, through the incorporation of gated units and nonlinear neurons into the storage units, coupled with the integration of vertical memory state transitions, the depth of the network has significantly increased, resulting in the enhancement of the sequence modeling capabilities. In terms of reliability of prediction results: The acceptable threshold for prediction error was set at 1000 kN. According to the average prediction results of all samples in the test set, it was found that the deep temporal sequence model was unable to accurately predict support resistance ahead. While the deep spatiotemporal sequence model demonstrated a predictable sequence length of 8. According to the prediction results of all samples in the test set, the accurate prediction rate increased from 30 %∼35 % to 65 %∼85 % and the maximum MAE value decreased from 10000 ∼ 12500kN to 6500 ∼ 9000kN in more than 20,000 test samples. The PredRNN model with an input sequence length of 50 achieves an accurate prediction rate of 85.48 % and a maximum MAE value of 6698kN, making it the most reliable prediction model. [ABSTRACT FROM AUTHOR]

Deep fuzzy systems are widely used in time series forecasting tasks due to their excellent nonlinear transformation capabilities and interpretability. However, traditional optimization methods and fixed network structures cannot make the deep fuzzy system obtain the expected prediction accuracy and generalization ability. Therefore, a novel hybrid deep fuzzy model (HDFM) is proposed in this paper. Firstly, two types fuzzy modules, namely the type-1 TSK fuzzy module (T1TSKFM) and the interval type-2 TSK fuzzy module (IT2TSKFM), are respectively presented and designed in detail. And, the gradient expressions of the fuzzy parameters, including the antecedent and consequent parameters, are also derived in detail. Secondly, in order to optimize the fuzzy parameters and to further accelerate the module convergence, a novel parameter optimization strategy is presented, combining the gradient descent with the Regularization, the DropRule and the AdaBound algorithms. Thirdly, a new stacked hybrid deep fuzzy architecture is proposed, which can be automatically trained and constructed using the designed T1TSKFM and the IT2TSKFM. Then, the detailed data-driven learning and updating strategy are given in step by step way. In addition, both the layered structure interpretability and the fuzzy rule interpretability are respectively analyzed. This can guarantee that the proposed model not only has the better forecasting accuracy, but also has the higher interpretability. Finally, in order to verify the effectiveness of the proposed method, several comparative experiments are given. Experimental results show that the forecasting performance of the proposed model outperforms the other comparisons, such as the DIRM-DFM, the IT2DIRM-DFM, the DCFS, and the ANFIS method. At the same time, the proposed model has better interpretability and more flexible construction property. • Giving the gradient formulas of the interval type-2 TSK fuzzy module. • Presenting a novel gradient-based optimization algorithm. • Proposing a new stacked hybrid deep fuzzy architecture. • Giving the detailed data-driven learning strategy for the proposed model. • Analyzing the interpretability in detail. [ABSTRACT FROM AUTHOR]

The application of Machine learning and deep learning techniques for time series forecasting has gained significant attention in recent years. Numerous endeavors have been devoted to automating forecasting through the utilization of cutting-edge neural networks. Notably, the recurrent neural network (LSTM – Long Short-Term Memory) has emerged as a central concept in most research endeavors. Although LSTM was initially introduced in 1997 for sequence modeling, subsequent updates have primarily focused on language learning tasks. These updates have introduced various computational mechanisms within the LSTM cell, including the forget gate, input gate, and output gate. In this study, we investigate the impact of each computational component in isolation to analyze their effects on time series forecasting tasks. Our experiments utilize the Jena weather dataset and Appliance Energy Usage time series for evaluation. The experimental results reveal that variations of the LSTM model outperform the most popular LSRM cell format in terms of error rate and training time. Specifically, the variations identified in this paper demonstrate superior generalization capabilities and yield reduced forecasting errors. [ABSTRACT FROM AUTHOR]

Accurate forecasting of the agricultural power load is important for rural electricity networks' safe and stable operation. But it is more uncertain and difficult to forecast than industrial, commercial, and residential loads. Multivariate time series prediction methods have problems such as poor refinement of weather services and difficulty in quantifying the influencing factors. Univariate time series prediction methods should be emphasized, and the characteristic information in the historical load data should be fully exploited. Therefore, this paper systematically investigates the decomposition strategy and attention-based recurrent neural network for univariate time-series multi-step prediction. The decomposition strategy is employed to generate derived variable sequences from the target sequence as inputs for the model, while the dual-stage attention mechanism is used for dynamically selecting features. A one-year real-world dataset of agricultural power loads in southern China is used for validation. The effectiveness of two different decomposition algorithms on dual-stage attention-based recurrent neural network (DARNN) is discussed, then the proposed ensemble empirical mode decomposition and dual-stage attention-based recurrent neural network (EEMD-DARNN) is compared with four univariate time series prediction models and three ablation models. The proposed model is optimal in four evaluation indicators, with R 2 , R M S E , M A P E , and M A E of 0.970, 12.298, 0.040, and 7.335 respectively in the shortest forecast horizon (15 min), and 0.941, 17.305, 0.065 and 11.347 respectively in the longest forecast horizon (1 h). It is a significant improvement over the existing models. In conclusion, EEMD-DARNN can achieve more accurate ultra-short-term agricultural power load forecasting. [ABSTRACT FROM AUTHOR]

Self-supervised graph representation Learning (SSGRL) is an emerging technique for machine learning-based expert applications. SSGRL can effectively encode unlabeled data into machine-understandable knowledge embeddings that can be used in downstream expert tasks. Recently, Masked-Graph-Model (MGM) has achieved extraordinary performances in the field of SSGRL. However, MGM inevitably suffers from a so-called Negative-Migration (NM) problem on unbalanced datasets for specific tasks such as graph classification. The NM problem forces the MGM to predict (reconstruct) the dominant nodes causing the ignoring of the minority but critical nodes. To this end, a novel online prediction module is proposed to dynamically predict the nodes with extended contexts that enable minority nodes can have more weights in the model. Based on the online prediction module, a graph representation learning model is proposed to alleviate the NM problem. Extensive experiments demonstrate that the model outperforms state-of-the-art methods on unbalanced datasets and achieves competitive performance on balanced datasets. Moreover, extra experiments demonstrate that the proposed method requires less computational resource overhead than existing methods. The code is available at: https://github.com/DuanhaoranCC/SimGOP. • This paper explores the negative migration problem encountered in unbalanced datasets. • We propose online predict contextual node features, but the not the original data. • The results of graph classification show that the proposed method achieves SOTA results. [ABSTRACT FROM AUTHOR]

Accurate and reliable load forecasting can ensure the safety and economy of power system operation. To improve the accuracy of short-term power load forecasting, this paper adopts feature extraction combined with a variety of machine learning methods,(Empirical wavelet decomposition (EWT), Convolutional neural network (CNN) model, Recurrent neural network (RNN) model, Long short-term memory network (LSTM) model, Bayesian optimization (BOA) algorithm). A new hybrid forecasting model, the EWT-CNN-S-RNN + LSTM model is proposed to predict power consumption. In this method, the fixed mode of EWT decomposition is used to extract statistical features, and the LSTM/RNN model is selected according to the statistical features. This method uses the Bayesian optimization (BOA) algorithm to optimize the parameters to solve the model gradient explosion problem. Based on the electricity consumption data of 2007 (Australian Energy Market Operator) and a total of 336 power load data, a good load forecasting result is obtained, which can provide indirect support for power regulation and sustainable development of electricity. [ABSTRACT FROM AUTHOR]

Persistent PM 2.5 pollution poses a serious threat to human health. Developing an accurate urban regional PM 2.5 forecasting is of practical significance for environmental protection. However, previous studies have mostly focused on individual monitoring stations, neglecting the influence of neighboring stations, which limits forecasting accuracy. Additionally, the PM 2.5 of a single monitoring station cannot reflect the overall situation of a region. Therefore, this paper develops a novel PM 2.5 spatiotemporal forecasting framework that combines graph convolutional module, temporal convolutional module, linear module. It enables the forecasting of PM 2.5 concentrations at multiple stations and multiple time steps in the future. Concretely, we utilize a mixed graph convolutional network to extract the spatial features of PM 2.5. Then, an improved temporal convolutional network, the second-order residual temporal convolutional network, is developed to capture complex temporal features. Following the classical "linear and non-linear" modeling strategy, a linear module is added to the forecasting framework. Experiments on the real air pollution dataset from Beijing demonstrate that our framework outperforms the state-of-the-art baselines. • Extracting and fusing spatiotemporal features of PM2.5 is crucial for prediction. • A novel spatiotemporal forecasting framework is proposed. • A second-order residual TCN is developed to extract complex temporal features. • Experiments on real air pollution dataset show our framework's superiority. [ABSTRACT FROM AUTHOR]

Compositional data (CoDa) is a type of relative data reflecting the proportion of different units in an integrated system, which can reveal the internal structure of observed objects and has been widely used to analyze practical economic and management problems. However, the summation constraint limits the prediction of compositional data time series. Especially, with small samples and poor information characteristics of the data set, the difficulties are further exacerbated. Compared with real-valued time series analysis, fuzzy time series (FTS) mainly transforms initial observed data to be different status so that the historical evolution rules can be derived. As a result, forms of initial observed data can be concealed. Correspondingly, this paper aims to establish a multi-rule combination prediction of compositional data based on a multivariate fuzzy time series model, named GA-CoDa-MFTS. Firstly, with the interaction of components, the adjustment rules between components are explored by using the fuzzy logical relationship (FLR). Secondly, the Genetic algorithm (GA) is used to find the optimal weight of the predicted values under different rules and derive the combined predicted values. Finally, the prediction effectiveness of the model is verified by CRMSE and CMAPE through a numerical study. • The fuzzy time series model and the compositional data are integrated. • The adjustment rules are explored by using the fuzzy logical relationship. • Genetic algorithm is used to get the optimal weight of different rules. • The combination model composed of different rules has better prediction effect. [ABSTRACT FROM AUTHOR]

With the emergence of artificial intelligence, deep learning techniques have been widely deployed in forecasting stock markets. However, existing deep-learning-based models for news-based forecasts of stock trends are mostly black-box and difficult to explain. The procedure by which how final predictions are made within models keeps unknown, making it hard to interpret why one prediction should be better than the other. To provide explanations on predictions, this paper proposes to inject causal inference into model procedures and causally interpret predictions. We first generate a causal graph from financial news, and then integrate the information in the causal graph into a neural network model for stock trend prediction. Moreover, in order to better extract keywords from financial news we introduce a novel keyword extraction method named Distinguishable Word Filtering by Kolmogorov–Smirnov Test (DWF-KST). The experiment results on five financial datasets demonstrate that not only our proposed model explicitly provides an interpretation of prediction results, but also outperforms the state-of-art methods. the achieved results boost predictions of S&P 500 2-category from 89.7% to 97.4%, 3-category from 77.4% to 82.5%, and 5-category from 61.5% to 71.6%. For the other two indexes, the performances of Dow index improve from 86.2% to 90.2% and Nasdaq index improve from 76.4% to 78.9%. • We propose a method for stock trend prediction. • We use causal inference based on unstructured data. • Improve the interpretability of DL-based stock prediction models. • We propose DWF-KST for keyword selection in financial news. [ABSTRACT FROM AUTHOR]

Student performance prediction is a critical task in supporting decision-making for Intelligent Tutoring Systems (ITS). Correct predictions of student performance are prerequisites for ITS to supply intelligent services and optimize learning efficiency, e.g., recommending the most appropriate learning resources for each student. Existing methods mainly include cognitive diagnosis and knowledge tracing, both of which focus on students' cognitive modeling based on their interactions on a sequence of items and give predictions by assessing if their cognitive states can meet the item requirements. Specifically, cognitive diagnosis only considers students' global static cognitive states, while knowledge tracing focuses on students' local dynamics. However, both global and local features are critical for predicting student performance. Therefore, in this paper, we propose a novel G lobal and L ocal N eural C ognitive (GLNC) model to capture both global and local features in student-item interactions for more accurate predictions. Specifically, we first learn students' global cognitive level according to all student-item interactions. Then, we propose a self-attentive encoder based on the scaled dot-product attention mechanism to extract the local cognitive dynamics and the dependencies between students' recent interactions. Finally, to obtain better predictions, we design a fused gate based on the similarity between students' recently responded items and the item to be predicted to adaptively combine the global and local features. To evaluate the effectiveness of GLNC, we compare it with both cognitive diagnosis and knowledge tracing methods. All experiments are conducted on three public datasets that contain real student-item interactions on mathematics or language courses from various ITS. The experimental results demonstrate that GLNC achieves an average score of 0.7810 on the AUC metric, 0.7627 on the ACC metric, 0.3987 on the RMSE metric, 0.2023 on the r 2 metric, respectively achieving an average improvement of 1.84%, 1.07%, 1.87%, and 11.38% in contrast to existing state-of-the-art methods (i.e., NCD and LPKT). Moreover, we further analyze the performance of GLNC under different probabilities of guessing and slipping, the results indicate that GLNC is more robust against the influence of noisy data while considering both global and local features. Benefiting from the superior accuracy and stability, our proposed GLNC has a wide range of potential implications for ITS, which can be easily applied to improve students' learning efficiency and experience. • Cognitive diagnosis focuses on students' global static cognitive states. • Knowledge tracing measures students' local dynamics in learning. • It is critical to make trade-offs between global and local features. • Our method achieves more precise and robust results on performance prediction. [ABSTRACT FROM AUTHOR]

In this paper, we present an algorithm designed to automatically merge predictions from a collection of individual prediction methods coded in R. The algorithm employs varying weights and decision rules to ascertain the optimal amalgamation of these methods, with the aim of forecasting historical time series data while minimizing human intervention. The algorithm serves as an automated component within the artificial intelligence toolkit. The proposed algorithm (al), denoted as "alPCA" is founded on principal component analysis (PCA), hence the acronym. Commencing with 52 configurations of 11 distinct methods available in R, we calculate several loss functions: specifically, scaled Mean Absolute Percentage Error (sMAPE) and Mean Absolute Scaled Error (MASE) for both fitting (Training Phase) and prediction (Validation Phase), along with Root Mean Squared Error (RMSE) and Overall Weighted Average (OWA) solely for prediction (Validation Phase). We then employ PCA to reduce the error matrix derived from this data to one or two dimensions. Subsequently, the methods are ranked based on their proximity to the highest score. A probability distribution is fitted to this proximity metric, and utilizing the percentiles of these values, the optimal methods for combination are selected. We propose three categories of weights derived from the PCA scores, encompassing the fitting sMAPE (Training Phase) and the prediction sMAPE (Validation Phase), to facilitate the amalgamation process. This approach is applied to seven distinct univariate time series across diverse domains, including automobile sales, electricity production, and CO2 levels. Additionally, a set of 100 random monthly series from the M4 competition is included in the analysis. To assess the predictive precision of our algorithm, we compare its performance against three widely utilized combined prediction algorithms available in R. We evaluate the outcomes in Test Phase (unseen data) using four distinct loss functions and conduct a sensitivity analysis to gauge the algorithm's robustness and efficacy across various specifications. [ABSTRACT FROM AUTHOR]

Machine learning techniques have been widely used and are mostly performed by predicting point estimation. Nevertheless, there are various scenarios that require more information beyond only point estimation. Probabilistic prediction is a typical research topic that provides probability distribution while predicting, which addressed with much attention recently. Most of the previous works make an assumption on the probabilistic distribution to a certain extent, which may lead to potential errors in the subsequent decision-making procedure. In this paper, we propose a distribution-free method for regression problems on real-value response under the probabilistic prediction framework and present an effective boosting-based method to perform the training process. Moreover, we introduce a further improved method to accomplish an unbiased mean estimation of the target distribution. Thorough experiments on multiple benchmark data are conducted to demonstrate the effectiveness of the proposed method with regard to different measures. • A distribution-free framework for probabilistic forecasting. • Use a mixture of Gaussian distribution as the target distribution. • An effective boosting-based strategy. • Providing unbiased mean estimation. [ABSTRACT FROM AUTHOR]

High-frequency financial data is more difficult to predict than low-frequency data because it possesses nonlinearity, nonstationarity, higher volatility, and long memory and is frequently accompanied by the jump phenomena. In this paper, the nonparametric regression (NR) model based on kernel function is used to fit the nonlinear relationship between the nonstationary series Y t and its lagging series to model the trend of high frequency financial time series. Furthermore, the deep learning LSTM (long short-term memory) model is applied to capture the high volatility and frequent jumps of high frequency financial data and to improve the forecasting accuracy of the residual series. The results demonstrate that the hybrid NR and LSTM model has greatly improved the forecasting accuracy in several evaluation criteria. In comparison to NR, support vector machine (SVM), LSTM, ARIMA and NR-SVM models, the mean absolute error (MAE) of NR-LSTM has reduced by 89.78%, 97.85%, 86.48%, 32.47% and 89%, respectively. In addition, we have constructed the trading strategy for the Shanghai-Shenzhen 300 index by using the NR-LSTM model. The NR-LSTM model can continue to provide good returns even during a bear market, which can serve as a guide for investors. Furthermore, the NR-LSTM model also exhibits the best forecasting effect when we model the high-frequency data of Ping An bank in China, the FTSE 100 index in the UK, and the S&P 500 index in the US. [ABSTRACT FROM AUTHOR]

In the context of a resilient energy system, accurate residential load forecasting has become a non-trivial requirement for ensuring effective management and planning strategy/policy development. Due to the highly stochastic nature of energy load profiles, it is difficult to predict accurately, and usually, predictions are error-prone. This paper explores the potential of Empirical Mode Decomposition (EMD) in simplifying the dynamics of complex demand profiles. The simplified components are then embedded within a deep learning model, specifically Convolution Neural Network (CNN) and Long Short-Term Memory (LSTM), to forecast short-term residential loads. The novel modelling framework integrates Bayesian optimisation strategy, feature decomposition technique, feature engineering phase, and percentile-based bias correction algorithm to enhance model accuracy. The model is developed using a case-study residential dwelling located in Fintry (Scotland), and the model performance is assessed over four forecast horizons. The overall efficiency of framework is also investigated for three algorithms: random forest, gradient boosting decision trees (GBDT), and an LSTM network. While EMD and feature engineering were found to greatly improve prediction accuracy, the number of IMFs used was shown to significantly impact the model's performance and computational complexity. The model was tested on two further case studies from Fintry. • A novel forecasting framework is proposed for short-term energy demand prediction. • Multi-domain features are extracted to increase the accuracy. • Utilising empirical mode decomposition to enhance framework performance. • Exhibit strong forecasting capability without the additional computational cost. • Excellent forecasting accuracy and model robustness are validated using three households. [ABSTRACT FROM AUTHOR]

Trajectory prediction aims to estimate the future location of the vehicle based on its historical motion state, which is essential for driving decision-making and local motion planning of smart vehicles. However, affected by the multiple complex interaction in the traffic scene, predicting future trajectory accurately is a challenging task. The majority of existing methods only focus on modeling the inter-vehicle interaction, while ignoring the influence of road alignment and driver's lane-change intention, making the poor performance of models, especially for long-term prediction or when the vehicle maneuvers laterally. To overcome the deficiencies, this paper proposes Intention-convolution and Hybrid-Attention Network (IH-Net) for reliable trajectory prediction. Specifically, we analyze the correlation of lane-change behavior and the motion state of the vehicle, and then the Intention-convolutional Social Pooling module (I-CS) is introduced to extract complete interaction including the driver's lane-change intention and inter-vehicle interaction. In addition, we construct a novel Hybrid Attention Mechanism (H-AM) to explore the trajectory periodicity formed under the restriction of road alignment, as well as the impacts of different features on trajectory prediction, which is used to improve the model's long-term prediction capacity. The model's prediction accuracy with RMSE loss function is tested on two public datasets NGSIM and highD, and the results demonstrate that IH-Net remarkably outperforms the state-of-art methods in long-term prediction. [ABSTRACT FROM AUTHOR]

Congestion is a bane of urban life that affects a large share of the population on a daily basis. Thus, congestion gets tremendous attention from city stakeholders, residents, and researchers. The key challenge to preventing congestion is to accurately predict the traffic status (e.g., speed) of a particular road segment which is greatly affected by many factors, such as spatial, temporal, and road conditions. Although several research studies have focused on preventing congestion, most prediction-based literature came short of accurate predictions regarding precision and time efficiency regarding large-scale datasets. This paper proposes a new hybrid approach called Grizzly. This approach utilizes an improved Sequence to Sequence Bi-directional Long Short Term Memory Neural Network model that integrates data pre-processing techniques such as normalization and embeddings to improve traffic prediction accuracy. Carried out experiments on two large-scale real-world datasets, namely PEMS-BAY and METR-LA, pinpointing that the proposed approach outperformed the pioneering competitors from time-series-based and hybrid neural network-based baselines in terms of the agreed-on evaluation criteria (precision and computation time). • Bidirectional LSTM-based approach has been developed for multistep speed prediction. • Seq-to-Seq architecture is relevant to capturing long-term temporal dependencies. • The combination of Embedding and Normalization improves prediction quality and cost. • Grizzly outperforms both classical baselines and hybrid Neural Network-based models. • Grizzly was evaluated on two large real-world datasets with three prediction tasks. [ABSTRACT FROM AUTHOR]

As changes in external weather introduce significant uncertainty into runoff prediction, probabilistic runoff forecasting becomes increasingly important for water resource managers. However, limited research has been conducted on the rational employment of decomposition techniques in probabilistic prediction, especially when considering external factors effectively. To fill the gap, this paper proposes an innovative framework for probabilistic runoff forecasting. It first utilizes the stepwise decomposition (SD) technique to obtain training and test samples. Then an integration structure considering external factors is designed to enhance the learning ability of the neural network for various features and decomposition sequences. The boundary correction process utilizes the dynamic time warping (DTW) method to optimize the boundaries of the obtained samples in SD. This comprehensive framework is named SDIFC. Building upon this framework, we achieve a synergistic integration of three distinct methods: seasonal-trend decomposition procedures based on loess (STL), least absolute shrinkage and selection operator (LASSO), and quantile regression long short-term memory (QRLSTM), named STL-LASSO-QRLSTM. Multiple experiments and evaluations demonstrate that the three methods are very well adapted to the SDIFC framework. The proposed forecasting method not only captures the runoff trend accurately but also quantifies the uncertainty in the prediction process. The proposed method exhibits significant advantages in probabilistic forecasting, achieving an average improvement of 28.1%, 35.5%, and 27.3% in continuous rank probability score (CRPS) when compared to the model without SDIFC, the model using SD in other frameworks, and the model without LASSO, respectively. [ABSTRACT FROM AUTHOR]

Accurate solar irradiance forecasting with fine spatiotemporal correlations is essential for photovoltaic power generation. However, current spatiotemporal methods are unable to extract spatial features appropriately, leading to rough prediction results. Consequently, a deep graph neural network is proposed in this paper for multi-site solar irradiance forecasting. Self-adaptive adjacency matrices were developed to better capture the intrinsic influence between sites under different climate types at various time steps. Based on the adaptive graph convolution and long short-term memory network (AGCLSTM), the spatiotemporal features of each site were extracted to obtain prediction results for all the sites in the research area simultaneously. In the experiments conducted on the datasets of three climate types, the average root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R) of the proposed model are observed to be in a range of 47.519–54.854 W/m2, 18.968–24.426 W/m2, and 0.9737–0.9889, respectively. Compared to the benchmark models, the average RMSE and MAE reduce by 1.31%–10.86% and 3.75%–18.57%, respectively. Moreover, the proposed method proves to be effective for scenes of solar irradiance fluctuations with the help of self-adaptive adjacency matrices, which enhances its generalization ability and prediction accuracy for one-hour-ahead solar irradiance forecasting. [ABSTRACT FROM AUTHOR]

High-dose single exposures of long-wavelength ultraviolet-A (UV-A) radiation may trigger severe biological and skin tissue damage in humans and animals, as well as photosynthetic damage in plants. In humans, the highly abundant UV-A is also linked to an increased risk of skin cancer. This paper develops a new forecasting framework tailored for very short-term UV-A predictions using a Bayesian optimized ensemble Neural Basis Expansion Analysis for Interpretable Time Series (B-E-NBEATS) method, integrated with cloud cover predictors and the solar zenith angle. The design phase of the proposed model entails feature selection using an efficient neighborhood component analysis algorithm and hyperparameter tuning with a Bayesian optimizer. To further enhance the predictive performance and reliability, the B-E-NBEATS model is calibrated via uncertainty quantification by adopting an ensemble approach with single-point models trained on three key loss functions. The prescribed model generates interpretable forecasts of 20-minute ahead UV-A by empowering the ensemble members to learn seasonal and trend components in UV-A datasets. With respect to the four benchmark models, the statistical metrics and visual infographics of predicted and observed UV-A reveal a superior forecasting capability of the proposed B-E-NBEATS model. The model yields a comparatively high correlation coefficient of 0.910, 0.869, 0.933, and 0.900 for the spring, summer, autumn, and winter testing phases, respectively. The superior performance by the newly designed hybrid ensemble model ascertains its potential utility in UV-A monitoring and mitigating subsequent harmful exposure risk for the general public, and animal and plant life. [Display omitted] • An interpretable AI method is designed to predict solar ultraviolet-A radiation. • Deep learning ensemble approach is adopted to enhance forecasting performance. • The model is optimized with neighborhood component analysis and Bayesian optimizer. • Model calibration by quantifying uncertainty offers reliable ultraviolet-A forecasts. • The proposed model predicts cloud-affected solar ultraviolet-A with high accuracies. [ABSTRACT FROM AUTHOR]

In implant prosthesis treatment, the design of the surgical guide heavily relies on the manual location of the implant position, which is subjective and prone to doctor's experiences. When deep learning based methods has started to be applied to address this problem, the space between teeth are various and some of them might present similar texture characteristic with the actual implant region. Both problems make a big challenge for the implant position prediction. In this paper, we develop a two-stream implant position regression framework (TSIPR), which consists of an implant region detector (IRD) and a multi-scale patch embedding regression network (MSPENet), to address this issue. For the training of IRD, we extend the original annotation to provide additional supervisory information, which contains much more rich characteristic and do not introduce extra labeling costs. A multi-scale patch embedding module is designed for the MSPENet to adaptively extract features from the images with various tooth spacing. The global–local feature interaction block is designed to build the encoder of MSPENet, which combines the transformer and convolution for enriched feature representation. During inference, the RoI mask extracted from the IRD is used to refine the prediction results of the MSPENet. Extensive experiments on a dental implant dataset through five-fold cross-validation demonstrated that the proposed TSIPR achieves superior performance than existing methods. [ABSTRACT FROM AUTHOR]

With the increasing demand for Chinese cold chain logistics, accurate forecasting becomes significantly important for decision-making and planning, but the data scarcity is still a challenging problem for modeling method. This paper proposes a novel grey prediction model with bi-level structure to improve the performance of cold chain logistics demand forecasting. The model adapts to the data characteristics and generates corresponding background series to adjust the parameters. A differential evolutionary algorithm is employed to merge the time response functions generated from the bi-level structure. The proposed model extends the applicability compared with existing similar models and can provide higher accuracy for data sequences with a high developing rate. The model outperforms the classic grey model and its extensions in terms of forecasting accuracy. The trend of Chinese cold chain logistics demand from 2021 to 2025 is studied with the proposed model. The computational results show that the proposed grey prediction model is significantly better than other models. [ABSTRACT FROM AUTHOR]

As an epigenetic modification that plays an important role in modifying gene function and controlling gene expression during cell development, DNA N4-methylcytosine (4mC) is still lack of researching. It is therefore necessary to accurately predict the 4mC sites to make fully aware of its mechanism and function. In this paper, we propose a novel model which is called Structural Sparse Regularized Random Vector Functional Link Network (SSR-RVFL) for predicting 4mC sites. Compared with other state-of-the-art methods, SSR-RVFL performs better and achieves higher prediction accuracy. There are total six benchmark datasets used in the experiments, namely C. e l e g a n s , D. e l a n o g a s t e r , E. c o l i , A. t h a l i a n a G. s u b t e r r a n e u s and G. p i c k e r i n g i i. Our model improves the accuracy by 0.42%, 0.45%, 0.48%, 0.91%, 0.66% and 0.7% on these six benchmark datasets respectively, so it can be regarded as a more effective prediction tool. • Structural Sparse Regularized Random Vector Functional Link Network (SSR-RVFL) is proposed for predicting 4mC sites. • SSR-RVFL performs better and achieves higher prediction accuracy than state-of-the-art methods. • An efficient iterative solution method for the proposed algorithm is provided. [ABSTRACT FROM AUTHOR]