Time Series Prediction Research Articles

Optimised Deep Learning for Time-Critical Load Forecasting Using LSTM and Modified Particle Swarm Optimisation

Short-term electric load forecasting is critical for power system planning and operations due to demand fluctuations driven by variable energy resources. While deep learning-based forecasting models have shown strong performance, time-sensitive applications require improvements in both accuracy and convergence speed. To address this, we propose a hybrid model that combines long short-term memory (LSTM) with a modified particle swarm optimisation (mPSO) algorithm. Although LSTM is effective for nonlinear time-series predictions, its computational complexity increases with parameter variations. To overcome this, mPSO is used for parameter tuning, ensuring accurate forecasting while avoiding local optima. Additionally, XGBoost and decision tree filtering algorithms are incorporated to reduce dimensionality and prevent overfitting. Unlike existing models that focus mainly on accuracy, our framework optimises accuracy, stability, and convergence rate simultaneously. The model was tested on real hourly load data from New South Wales and Victoria, significantly outperforming benchmark models such as ENN, LSTM, GA-LSTM, and PSO-LSTM. For NSW, the proposed model reduced MSE by 91.91%, RMSE by 94.89%, and MAPE by 74.29%. In VIC, MSE decreased by 91.33%, RMSE by 95.73%, and MAPE by 72.06%, showcasing superior performance across all metrics.

An Efficient Deep Learning Framework for Optimized Event Forecasting

There have been several catastrophic events that have impacted multiple economies and resulted in thousands of fatalities, and violence has generated a severe political and financial crisis. Multiple studies have been centered around the artificial intelligence (AI) and machine learning (ML) approaches that are most widely used in practice to detect or forecast violent activities. However, machine learning algorithms become less accurate in identifying and forecasting violent activity as data volume and complexity increase. For the prediction of future events, we propose a hybrid deep learning (DL)-based model that is composed of a convolutional neural network (CNN), long short-term memory (LSTM), and an attention layer to learn temporal features from the benchmark the Global Terrorism Database (GTD). The GTD is an internationally recognized database that includes around 190,000 violent events and occurrences worldwide from 1970 to 2020. We took into account two factors for this experimental work: the type of event and the type of object used. The LSTM model takes these complex feature extractions from the CNN first to determine the chronological link between data points, whereas the attention model is used for the time series prediction of an event. The results show that the proposed model achieved good accuracies for both cases—type of event and type of object—compared to benchmark studies using the same dataset (98.1% and 97.6%, respectively).

Improving airborne laser scanning-based species-specific forest volume estimation using sentinel-2 time series

ABSTRACT Species-specific timber volume estimates are required to support forest planning and conservation. We evaluated whether additional predictors from a Sentinel-2 time series can improve airborne laser scanning (ALS)-based estimation of species-specific timber volumes. Furthermore, we determined the satellite image dates that provided the greatest improvement in accuracy. The Sentinel-2 time series was constructed to cover 1 March–30 November time-period, with a focus on late spring and early summer. The estimation was done using the k Most Similar Neighbor method and predictors extracted from the ALS data and Sentinel-2 images. Our best model included both ALS and Sentinel-2 time-series predictors, and the relative root mean square error (RMSE) values for pine, spruce and deciduous timber volumes were 40.8%, 57.0% and 51.3%, respectively (mean 49.7%). All deciduous trees were treated as one species. When bands from an individual image were used instead of the time series, the best result was obtained with an image from September where the respective relative RMSE values were 42.2% (deciduous), 58.4% (pine) and 60.6% (spruce) with a mean value of 53.7%. A fusion of a Sentinel-2 time-series and ALS data can improve species-specific estimation results compared to the use of individual Sentinel-2 images or ALS only.

Land Subsidence Predictions Based on a Multi-Component Temporal Convolutional Gated Recurrent Unit Model in Kunming City

Land subsidence (LS) is a geological hazard driven by both natural conditions and human activities. Traditional LS time-series prediction models often struggle to accurately capture nonlinear data characteristics, leading to suboptimal predictions. To address this issue, this paper introduces a multi-component temporal convolutional gate recurrent unit (MC-TCGRU) model, which integrates a fully adaptive noise-ensemble empirical-mode decomposition algorithm with a deep neural network to account for the complexity of time-series data. The model was validated using typical InSAR subsidence data from Kunming, analyzing the impact of each component on the prediction performance. A comparative analysis with the TCGRU model and models based on seasonal-trend decomposition using LOESS (STL) and empirical-mode decomposition (EMD) revealed that the MC-TCGRU model significantly enhanced the prediction accuracy by reducing the complexity of the original data. The model achieved R² values of 0.90, 0.93, 0.51, 0.93, and 0.96 across five points, outperforming the compared models.

ATPAM: Adversarial Temporal Pattern Attention Mechanism

Abstract. Time series forecasting has a large number of applications in daily life, for instance the prediction of stock prices, electricity consumption, exchange rate changes etc. However, the existing time series prediction methods have limitations. The most significant one is that when all the prediction models get the predicted value, a new round of iteration starts after the loss function is calculated with the corresponding real data. This may cause the accumulation of errors, since there is only one loss function to measure the difference between the predicted value and the ground truth, it will make the connection weak and mainly depend on the accuracy of the prediction model. Unfortunately, there are few time series prediction models with high accuracy in reality. To solve the issue, in this paper, we propose a new time series forecasting model Adversarial Temporal Pattern Attention Mechanism (ATPAM), which based on Generative Adversarial Nets (GANs). ATPAM adopts a Temporal Pattern Attention model as the generator to learn time-invariant temporal patterns, and use a discriminator to improve the prediction performance and do auxiliary adversarial training. Extensive experiments on several real-world datasets show the effectiveness of our method.

Multi-criteria Forecast Combination Method with Nonlinear Programming for time series prediction models

Multi-criteria Forecast Combination Method with Nonlinear Programming for time series prediction models

Research on the climate change and prediction of the temperature in Beijing

Abstract. Over the past few years, global warming has become a growing threat to the human living environment. Thus, preventing extreme temperature changes is a crucial area of study for the global community. However, temperature changes have diverse causes, making it hard to accurately calculate the data that impacts subsequent temperature forecasts. The ARIMA model is widely favoured by researchers as a typical time series prediction model. Additionally, it holds significant research value in forecasting dynamic models. This paper will use the ARIMA model to forecast temperature changes in Beijing for the next month, starting on 21 July. The findings show that the temperature remains relatively stable in the upcoming month, with minimal fluctuations. Based on historical data, this is the peak temperature in Beijing for the year. So, if the temperatures are not extremely high now, they might not get higher later. This forecast can be used as a guide for managing extreme temperatures in Beijing in the future.

Utilising ARIMA models to predict the mass balance of the Antarctic ice sheet and comprehend its fluctuations

Abstract. The mass balance of the Antarctic Ice Sheet is forecasted in this study using ARIMA models, with a particular emphasis on the distinct regions of East and West Antarctica. The research endeavours to forecast the trends in ice mass balance over the next 12 years by examining data from 1992 to 2021, thereby elucidating the potential impacts of climate change on these critical regions. The ARIMA model, which is renowned for its capacity to capture and predict time series data, has identified substantial trends that indicate a persistent loss of ice mass, particularly in West Antarctica, which has experienced substantial declines in recent decades. The model's predictions suggest that, despite the potential for a reduction in the rate of loss, the overall trend is consistent with the ongoing reduction of ice mass. These results underscore the pressing necessity for ongoing research and monitoring to gain a more comprehensive understanding of the consequences of these developments. The study also emphasises the necessity of enhancing predictive models by incorporating supplementary environmental variables to provide more comprehensive insights into the future of the Antarctic Ice Sheet and its global impacts, thereby increasing their accuracy.

AutoML and Automated Data Science by Democratizing AI through End-to-End Automation

The rise of data-driven decision-making has led to a significant demand for data science and machine learning (ML) solutions across industries. However, developing these solutions requires extensive expertise in data preprocessing, feature engineering, model selection, hyperparameter tuning and evaluation. AutoML (Automated Machine Learning) and Automated Data Science (AutoDS) have emerged as transformative approaches that aim to democratize data science by automating the endto-end ML pipeline. This paper explores the foundational concepts of AutoML, highlighting key techniques and algorithms, such as neural architecture search (NAS), hyperparameter optimization and meta-learning. We delve into AutoDS's broader scope, which seeks to fully automate tasks from data acquisition to deployment. Real-world applications, such as predictive modeling, anomaly detection and time series forecasting, are examined to demonstrate the impact of these technologies. Additionally, the paper analyzes the current frameworks and platforms facilitating automation, including Auto-sklearn, Google AutoML and H2O.ai and evaluates their performance across different tasks. While the potential to accelerate data science workflows and make AI accessible to non-experts is evident, challenges remain, particularly regarding transparency, interpretability and ethical considerations in fully automated systems. This research provides insights into current trends, future opportunities and the transformative role of AutoML and AutoDS in driving innovation in the data science landscape.

Dynamic Black–Litterman Portfolios Incorporating Asymmetric Fractal Uncertainty

This study investigates the profitability of portfolios that integrate asymmetric fractality within the Black–Litterman (BL) framework. It predicts 10-day-ahead exchange-traded fund (ETF) prices using recurrent neural networks (RNNs) based on historical price information and technical indicators; these predictions are utilized as BL views. While constructing the BL portfolio, the Hurst exponent obtained from the asymmetric multifractal detrended fluctuation analysis is employed to determine the uncertainty associated with the views. The Hurst exponent describes the long-range persistence in time-series data, which can also be interpreted as the uncertainty in time-series predictions. Additionally, uncertainty is measured using asymmetric fractality to account for the financial time series’ asymmetric characteristics. Then, backtesting is conducted on portfolios comprising 10 countries’ ETFs, rebalanced on a 10-day basis. While benchmarking to a Markowitz portfolio and the MSCI world index, profitability is assessed using the Sharpe ratio, maximum drawdown, and sub-period analysis. The results reveal that the proposed model enhances the overall portfolio return and demonstrates particularly strong performance during negative trends. Moreover, it identifies ongoing investment opportunities, even in recent periods. These findings underscore the potential of fractality in adjusting uncertainty for diverse portfolio optimization applications.

Containerized Freight Index Prediction Integrating Deep Learning and Time Series Analysis

In the context of globalization, the shipping industry, as a core pillar of international trade, is crucial for global economic stability. This study focuses on the China Containerized Freight Index (CCFI) and proposes a composite deep learning model combining CEEMDAN, CNN, LSTM, and SENet for CCFI analysis and prediction. Initially, the CEEMDAN algorithm is used to decompose the CCFI time series, extracting useful signals from noise. Then, CNN extracts features, LSTM model sequences, and SENet enhances feature representation. The CEEMDAN-CNN-LSTM-SENet model achieves superior predictive performance with MAE of 29.5171, RMSE of 33.7638, and MAPE of 1.56%. Compared to other models like ECL (MAE 77.0951, RMSE 57.3559, MAPE 3.25%), our model shows approximately 61.7% improvement in MAE. This research offers a robust tool for shipping market analysis and policy-making, providing new perspectives for handling complex economic data.

Enhancing Daily Crash Count Prediction using Deep Learning with Window Size Selection and Seasonality Predictor Integration

The escalating need for proactive safety measures, coupled with advancements in data collection and analytical techniques, has significantly refined the accuracy of crash count predictions, shifting from annual scales to finer daily or hourly estimates. This research places emphasis on recurrent neural networks (RNNs), specifically the long short-term memory (LSTM) model, acknowledged for effectively managing sequential data in time series predictions. Paramount considerations encompass the treatment of input data, including the decision to incorporate temporal features alongside endogenous historical target values, and the establishment of an optimal window size for data input. Despite the critical nature of these facets, exhaustive studies concurrently investigating both under controlled conditions are scarce. This research addresses this gap, assessing diverse scenarios featuring distinct temporal treatments and window sizes, and employing an LSTM model with uniform fine-tuned parameters to ensure a fair comparison. Findings indicate a significant variation in performance among models employing different window sizes and month predictor integration, under identical RNN structures and LSTM configurations. The best model outperformed the least effective by roughly 30% concerning root mean square error (RMSE) and ranking correlation between predicted and actual target crash counts in the test datasets. Unlike seasonality predictor treatment, diverse window size selections did not lead to statistically significant differences in model performance. Generally, a window size of 3 performed worst under most conditions, followed by a size of 14. Sizes of 7 and 28 performed best in nearly an equal numbers of cases.

Comparison of transformer, LSTM and coupled algorithms for soil moisture prediction in shallow-groundwater-level areas with interpretability analysis

Accurate quantification of soil moisture is essential for understanding water and energy exchanges between the atmosphere and the Earth’s surface, as well as for agricultural applications. Predicting soil moisture content is vital for efficient water management, irrigation scheduling, and drought monitoring. Traditional forecasting methods, such as numerical regression models, often struggle due to various influencing factors and poor observation data quality. In contrast, deep learning algorithms, particularly recurrent and convolutional neural networks, show promise in predicting nonlinear data like soil moisture. This study focuses on shallow groundwater regions, using groundwater levels and meteorological data as features while coupling the Transformer model with other neural network structures. We investigate the potential of attention-based neural networks for soil moisture time series prediction. Our findings demonstrate that the Transformer model achieves an average R2 of 0.523 across different time lags, outperforming the LSTM model with an R2 of 0.485. The introduction of LSTM enhances the Transformer’s stability in handling temporal changes. Additionally, we verified the importance of groundwater for soil moisture prediction. This study introduces new methods for soil moisture prediction and offers new insights and recommendations for the development of artificial intelligence technology for soil moisture prediction.

A New Algorithm for Predicting Dam Deformation Using Grey Wolf-Optimized Variational Mode Long Short-Term Neural Network

To solve the problems of difficult-to-model parameter selections, useful-signal extraction and improper-signal decomposition in nonlinear, non-stationary dam displacement time series prediction methods, we propose a new predictive model for grey wolf optimization and variational mode decomposition and long short-term memory (GVLSTM). Firstly, we used the grey wolf optimization (GWO) algorithm to optimize the parameters of variable mode decomposition (VMD), obtaining the optimal parameter combination. Secondly, we used multiscale permutation entropy (MPE) as a standard to select signal screening, determining and recon-structing the effective modal components. Finally, the long short-term memory neural network (LSTM) was used to learn the dam deformation characteristics. The result shows that the GVLSTM model can effectively reduce the estimation deviation of the prediction model. Compared with VMDGRU and VMDANN, the average RMSE and MAE value of each station is increased by 19.11%~28.58% and 27.66%~29.63%, respectively. We used determination (R2) coefficient to judge the performance of the prediction model, and the value of R2 was 0.95~0.97, indicating that our method has good performance in predicting dam deformation. The proposed method has outstanding advantages of high accuracy, reliability, and stability for dam deformation prediction.

Time-series water prediction based on feature clustering fusion and XGboost

Abstract. The time series prediction problem has a very wide range of applications in many fields. Most scholars use the LSTM class of algorithms for prediction. However, this method consumes a significant amount of computational power and presents several challenges. To address this issue, this paper proposes a time series forecasting method based on feature fusion and XGBoost. Specifically, we first utilize holiday information and the K-Means algorithm for feature extraction to expand the feature dimensions of the dataset, and then employ XGBoost as a model for training and prediction. Experiments demonstrate that the method proposed in this paper significantly reduces error compared to other traditional machine learning and deep learning methods, while the training time is much shorter than these methods. For example, compared with LSTM, the MSLE of this model decreases by 1.42%, while the training time is only 0.15% of that of LSTM. This greatly saves on training costs and computational power consumption. This confirms the effectiveness of using machine learning and clustering algorithms in time series prediction and provides new methods and practical application directions for future time series prediction models.

Study of Five-Hundred-Meter Aperture Spherical Telescope Feed Cabin Time-Series Prediction Studies Based on Long Short-Term Memory–Self-Attention

The Five-hundred-meter Aperture Spherical Telescope (FAST), as the world’s most sensitive single-dish radio telescope, necessitates highly accurate positioning of its feed cabin to utilize its full observational potential. Traditional positioning methods that rely on GNSS and IMU, integrated with TS devices, but the GNSS and TS devices are vulnerable to other signal and environmental disruptions, which can significantly diminish position accuracy and even cause observation to stop. To address these challenges, this study introduces a novel time-series prediction model that integrates Long Short-Term Memory (LSTM) networks with a Self-Attention mechanism. This model can hold the precision of feed cabin positioning when the measure devices fail. Experimental results show that our LSTM-Self-Attention model achieves a Mean Absolute Error (MAE) of less than 10 mm and a Root Mean Square Error (RMSE) of approximately 12 mm, with the errors across different axes following a near-normal distribution. This performance meets the FAST measurement precision requirement of 15 mm, a standard derived from engineering practices where measurement accuracy is set at one-third of the control accuracy, which is around 48 mm (according to the accuracy form the official threshold analysis on the focus cabin of FAST). This result not only compensates for the shortcomings of traditional methods in consistently solving feed cabin positioning, but also demonstrates the model’s ability to handle complex time-series data under specific conditions, such as sensor failures, thus providing a reliable tool for the stable operation of highly sensitive astronomical observations.

Meta-heuristics time series prediction with Reduced Noise and Time Series Disintegration in Remote Sensing

Meta-heuristics time series prediction with Reduced Noise and Time Series Disintegration in Remote Sensing

Multi-task photonic reservoir computing: wavelength division multiplexing for parallel computing with a silicon microring resonator

Nowadays, as the ever-increasing demand for more powerful computing resources continues, alternative advanced computing paradigms are under extensive investigation. Significant effort has been made to deviate from conventional Von Neumann architectures. In-memory computing has emerged in the field of electronics as a possible solution to the infamous bottleneck between memory and computing processors, which reduces the effective throughput of data. In photonics, novel schemes attempt to collocate the computing processor and memory in a single device. Photonics offers the flexibility of multiplexing streams of data not only spatially and in time, but also in frequency or, equivalently, in wavelength, which makes it highly suitable for parallel computing. Here, we numerically show the use of time and wavelength division multiplexing (WDM) to solve four independent tasks at the same time in a single photonic chip, serving as a proof of concept for our proposal. The system is a time-delay reservoir computing (TDRC) based on a microring resonator (MRR). The addressed tasks cover different applications: Time-series prediction, waveform signal classification, wireless channel equalization, and radar signal prediction. The system is also tested for simultaneous computing of up to 10 instances of the same task, exhibiting excellent performance. The footprint of the system is reduced by using time-division multiplexing of the nodes that act as the neurons of the studied neural network scheme. WDM is used for the parallelization of wavelength channels, each addressing a single task. By adjusting the input power and frequency of each optical channel, we can achieve levels of performance for each of the tasks that are comparable to those quoted in state-of-the-art reports focusing on single-task operation. We also quantify the memory capacity and nonlinearity of each parallelized RC and relate these properties to the performance of each task. Finally, we provide insight into the impact of the feedback mechanism on the performance of the system.

Analysis and modelling of global online public interest in multiple other infectious diseases due to the COVID-19 pandemic.

Previous research has demonstrated the applicability of Google Trends in predicting infectious diseases. This study aimed to analyze public interest in other infectious diseases before and after the outbreak of COVID-19 via Google Trends data and to predict these trends via time series models. Google Trends data for 12 common infectious diseases were obtained in this study, covering the period from 1 February 2018 to 5 May 2023. The ARIMA, TimeGPT, XGBoost, and LSTM algorithms were then utilized to establish time series prediction models. Our study revealed a significant decrease in public interest in most infectious diseases at the beginning of the pandemic outbreak, followed by a rebound in the post-pandemic era, which is consistent with reported disease incidences. Furthermore, our prediction models demonstrated good accuracy, with TimeGPT showing unique advantages. Our study highlights the potential application value of Google Trends and large pre-trained models for infectious disease prediction.

Prediction model of permeability index for blast furnace based on WD-NL-transformer

The permeability index is an important variable to reflect the operating state of a blast furnace quickly, intuitively and comprehensively. Given the multi-scale, nonlinear and massive characteristics of the blast furnace smelting process data, a prediction model of permeability index based on the WD-NL-transformer (wavelet de-noising-nonlinear-transformer) method was proposed. The key variables affecting permeability index were selected from multiple perspectives by Pearson correlation coefficient and copula entropy. The extreme outliers were optimised using box plots and Lagrange linear interpolation. A data set was constructed after wavelet de-noising to build the prediction model with nonlinear processing capability, time-series prediction capability, and early stopping method. The results show that the MSE of WD-NL-transformer is 0.0093, with the RMSE of 0.0086. Meanwhile, the R2 is 0.9859, and the running time is 19.59 s, which means the model has great prediction accuracy and inference speed. The proposed model could provide theoretical support for advanced control of permeability index, which is of practical significance to ensure the stability of blast furnace smelting and accelerate the metallurgical process of intelligent automation.