Time Series Prediction Model Research Articles

Reduced-order modeling of unsteady fluid flow using neural network ensembles

The use of deep learning has become increasingly popular in reduced-order models (ROMs) to obtain low-dimensional representations of full-order models. Convolutional autoencoders (CAEs) are often used to this end as they are adept at handling data that are spatially distributed, including solutions to partial differential equations. When applied to unsteady physics problems, ROMs also require a model for time-series prediction of the low-dimensional latent variables. Long short-term memory (LSTM) networks, a type of recurrent neural network useful for modeling sequential data, are frequently employed in data-driven ROMs for autoregressive time-series prediction. When making predictions at unseen design points over long time horizons, error propagation is a frequently encountered issue, where errors made early on can compound over time and lead to large inaccuracies. In this work, we propose using bagging, a commonly used ensemble learning technique, to develop a fully data-driven ROM framework referred to as the CAE-eLSTM ROM that uses CAEs for spatial reconstruction of the full-order model and LSTM ensembles for time-series prediction. When applied to two unsteady fluid dynamics problems, our results show that the presented framework effectively reduces error propagation and leads to more accurate time-series prediction of latent variables at unseen points.

화물차의 제동장치에서 발생하는 과열 예측방안 연구

Recently, due to the increase in domestic and international online e-commerce platforms and the increase in container traffic at domestic ports, the operating ratio of large trucks has increased, and the number of truck fires is continuously increasing. In particular, spontaneous combustion is the most common cause of truck fires. Various academic approaches have been attempted to prevent truck fires, but due to the lack of research on the spontaneous tire ignition phenomenon that occurs during braking, this research directly designed and manufactured an experimental device to establish an environment similar to the braking system of a truck. A non-contact temperature sensor was installed on the brake device of the experimental device to collect temperature data generated from the brake device. Based on the data collected from the temperature sensor of the brake device and the temperature sensor on the tire surface, the ARIMA model among the time series prediction models was used to Appropriate parameters were selected to suit the temperature change trend, and as a result of comparing and analyzing the measured and predicted data, an accuracy of over 90% was obtained. Based on this, a plan was proposed to reduce the rate of fires in trucks by providing real-time warnings and support for truck drivers to respond to overheating phenomena occurring in the braking system.

Comparative Study of Artificial Neural Networks and Hidden Markov Model for Financial Time Series Prediction

Financial Time Series analysis and prediction isone of the interesting areas in which past data could be usedto anticipate and predict data an d information about future.There are many artificial intelligence approaches used in theprediction of time series, such as Artificial Neural Networks (ANN) and Hidden Markov Models (HMM). In this paperHMM and HMM approaches for predicting financial timeseries are presented. ANN and HMM are used to predicttime series that consists of highest and lowest Forex indexseries as input variable. Both of ANN and HMM are trained on the past dataset of the chosen currencies (such as EURO/USD which is used in this paper). The trained ANN andHMM are used to search for the variable of interestbehavioral data pattern from the past dataset. The obtained results was compared with real values from Forex (ForeignExchange) market database [1]. The power and predictiveability of the two models are evaluated on the basis of MeanSquare Error (MSE). The Experimental results obtained areencouraging, and it demonstrate that ANN and HMM can closely predict the currency market, with a small differentin predicting performance

Enhancing Water Management: A Comparative Analysis of Time Series Prediction Models for Distributed Water Flow in Supply Networks

Enhancing Water Management: A Comparative Analysis of Time Series Prediction Models for Distributed Water Flow in Supply Networks

A long-term multivariate time series prediction model for dissolved oxygen

Accurate and efficient long-term prediction of marine dissolved oxygen (DO) is crucial for the sustainable development of aquaculture. However, the multidimensional time dependency and lag effects of marine chemical variables present significant challenges when handling multiple inputs in univariate long-term prediction tasks. To address these issues, we designed a multivariate time-series long-term prediction model (LMFormer) based on the Transformer architecture. The proposed multivariate time-series decomposition strategy effectively leverages the feature information of prediction variables at different scales, thereby reducing the loss of critical information. Additionally, a dynamic variable selection strategy based on a gating mechanism was designed to optimize the collinearity problem in the multivariate data feature extraction process. Finally, an efficient two-stage attention architecture is proposed to effectively capture the long-range dependencies between dynamic features. This study conducted high-precision 7-day advance DO long-term predictions in two case studies, the environmentally stable Shandong Peninsula in China and the San Juan Islands in the United States, which are affected by extreme conditions such as ocean currents. The experimental results demonstrate the superior prediction performance and generalizability of the designed model. In the Shandong Peninsula case, the mean absolute error (MAE), root mean square error (RMSE), coefficient of determination (R2), and Kling–Gupta efficiency (KGE) reached 0.0159, 0.126, 0.9743, and 0.9625, respectively. In the San Juan Islands case, the MAE was reduced by an average of 42.34% compared to that of the baseline model, the RMSE was reduced by an average of 24.57%, the R2 increased by 22.54%, and the KGE improved by an average of 12.04%. Overall, the proposed prediction model effectively achieves long-term prediction of multivariate marine chemical data, providing valuable references for sustainable management and decision-making in aquaculture.

Prediction of significant wave height using a VMD-LSTM-rolling model in the South Sea of China

Accurate prediction of significant wave height is crucial for ocean engineering. Traditional time series prediction models fail to achieve satisfactory results due to the non-stationarity of significant wave height. Decomposition algorithms are adopted to address the problem of non-stationarity, but the traditional direct decomposition method exists information leakage. In this study, a hybrid VMD-LSTM-rolling model is proposed for non-stationary wave height prediction. In this model, time series are generated by a rolling method, after which each time series is decomposed, trained and predicted, then the predictions of each time series are combined to generate the final prediction of significant wave height. The performance of the LSTM model, the VMD-LSTM-direct model and the VMD-LSTM-rolling model are compared in terms of multi-step prediction. It is found that the error of the VMD-LSTM-direct model and the VMD-LSTM-rolling model is lower than that of the LSTM model. Due to the decomposition of the testing set, the VMD-LSTM-direct model has a slightly higher accuracy than the VMD-LSTM-rolling model. However, given the issue of information leakage, the accuracy of the VMD-LSTM-direct model is considered false. Thus, it has been proved that the VMD-LSTM-rolling model exhibits superiority in predicting significant wave height and can be applied in practice.

BiLSTM-MLAM: A Multi-Scale Time Series Prediction Model for Sensor Data Based on Bi-LSTM and Local Attention Mechanisms.

This paper introduces BiLSTM-MLAM, a novel multi-scale time series prediction model. Initially, the approach utilizes bidirectional long short-term memory to capture information from both forward and backward directions in time series data. Subsequently, a multi-scale patch segmentation module generates various long sequences composed of equal-length segments, enabling the model to capture data patterns across multiple time scales by adjusting segment lengths. Finally, the local attention mechanism enhances feature extraction by accurately identifying and weighting important time segments, thereby strengthening the model's understanding of the local features of the time series, followed by feature fusion. The model demonstrates outstanding performance in time series prediction tasks by effectively capturing sequence information across various time scales. Experimental validation illustrates the superior performance of BiLSTM-MLAM compared to six baseline methods across multiple datasets. When predicting the remaining life of aircraft engines, BiLSTM-MLAM outperforms the best baseline model by 6.66% in RMSE and 11.50% in MAE. In the LTE dataset, it achieves RMSE improvements of 12.77% and MAE enhancements of 3.06%, while in the load dataset, it demonstrates RMSE enhancements of 17.96% and MAE improvements of 30.39%. Additionally, ablation experiments confirm the positive impact of each module on prediction accuracy. Through segment length parameter tuning experiments, combining different segment lengths has resulted in lower prediction errors, affirming the effectiveness of the multi-scale fusion strategy in enhancing prediction accuracy by integrating information from multiple time scales.

Predictive Root Based Bootstrap Prediction Intervals in Neural Network Models for Time Series Forecasting

Predictive Root Based Bootstrap Prediction Intervals in Neural Network Models for Time Series Forecasting

Reviewing Optimization Techniques in Supply Chains: AI and Blockchain Perspectives

The integration of Artificial Intelligence (AI) and Big Data Analytics (BDA) into Supply Chain Management (SCM) has transformed the industry by enhancing efficiency, accuracy, and responsiveness. This review paper provides a comprehensive analysis of current AI applications in SCM, focusing on demand forecasting, inventory management, logistics, and transportation. Various AI techniques, including time-series forecasting, clustering, neural networks, SARIMA, and LSTM models, are discussed in detail. The impact of cutting-edge technologies on supply chain traceability and efficiency, such as blockchain and the Internet of Things (IoT), is also examined in this article. Despite the significant advancements, challenges such as gaps in closed-loop supply chains, terminology inconsistencies, and the need for better technical-managerial alignment persist. This review recognizes future research directions to address and solve these challenges and highlights the potential for AI to drive further innovations in SCM. Through case studies and bibliometric analyses, this paper underscores the significance of a comprehensive strategy for supply chain redesign, integrating physical, facilities, and information management to enhance sustainability and market responsiveness. Keywords— Supply chain management , Artificial Intelligence (AI), Big Data Analytics (BDA), Demand forecasting, Inventory management, Logistics optimization, Blockchain , Internet of Things (IoT), Smart Transportation, Tactical planning, Strategic planning, Advanced available-to-promise (AATP), Sustainability in supply chain.

Comparative Analysis of Univariate SARIMA and Multivariate VAR Models for Time Series Forecasting: A Case Study of Climate Variables in Ninahvah City, Iraq

This study involves a comparison between the application of the univariate SARIMA model and the utilization of VAR methods (vector autoregressive models) for multivariate time series analysis. The analysis is conducted using three-time series variables derived from data representing the monthly average of Humidity (H), Rainfall (R), and Temperature (T) in Ninahvah City, Iraq. Both univariate and multivariate time series approaches are employed to model these series. The paper also outlines the implementation of vector autoregressive, structural vector autoregressive, and structural vector error correction models using the 'vars' package. Additionally, it provides functions for diagnostic testing, estimation of constrained models, prediction, causality analysis, impulse response analysis, and forecast error variance decomposition. Furthermore, it introduces three fundamental functions, VAR, SVAR, and SVEC, for estimating these models. The comparison between the methods is based on evaluating the mean error produced by each approach. The findings of the study indicate that univariate linear stationary methods outperform multivariate models. The analysis of the data was carried out using the R software platform. The primary objective is to assess the performance of univariate and multivariate time series models in handling the given data. The research gap lies in the need for a comparative evaluation of SARIMA and VAR methods for time series analysis in the context of monthly environmental variables. These models were chosen due to their effectiveness in capturing temporal dependencies and interactions among multiple variables in time series data, providing a comprehensive analysis of climatic patterns in Ninahvah City, Iraq. The study aims to address the research gap by comparing these models and justifying their selection based on their capabilities to analyze the specified time series data.

Forecasting and uncertainty analysis of tailings dam system safety based on data mining techniques

Tailings dams, as critical infrastructure, play a vital role in ensuring the safety and reliability of tailings pond systems. Predicting the trend of tailings dam monitoring parameters helps monitor its operational status and support safety management and decision-making. The traditional time series prediction model focuses on point prediction while neglecting the nonlinear feature of data and the uncertainty of prediction results, which is far from meeting the requirements for risk monitoring and safety management of tailings dam systems. To address this, Deep Auto-Encoder is used to extract and optimize features of multi-dimensional time series; Whale Optimization Algorithm, Long Short Term Memory, and Kernel Density Estimation are used to build a monitoring parameters trend interval prediction model. It can effectively predict and analyze the uncertainty of tailings dam monitoring parameters. Time series data of monitoring parameters are collected from a tailings dam in Anhui, China and modeled to verify the effectiveness of the model. The results show that the hybrid prediction model performs the best in monitoring parameter prediction accuracy and trend. The Root Mean Square Error values of the Whale Optimization Algorithm and Long Short Term Memory prediction for reservoir water level, surface displacement, internal displacement, phreatic line, and beach width are 0.0160, 1.0405, 0.0389, 0.0176, and 0.7525, with errors reduced by 0.37 % to 17.92 % compared to other models. The uncertainty analysis confirms the high reliability of the model. The Kernel Density Estimation effectively forecasts the fluctuation range of predicted values, significantly improving the prediction effectiveness of the monitoring parameters.

A Review of Time-Series Forecasting Algorithms for Industrial Manufacturing Systems

Time-series forecasting is crucial in the efficient operation and decision-making processes of various industrial systems. Accurately predicting future trends is essential for optimizing resources, production scheduling, and overall system performance. This comprehensive review examines time-series forecasting models and their applications across diverse industries. We discuss the fundamental principles, strengths, and weaknesses of traditional statistical methods such as Autoregressive Integrated Moving Average (ARIMA) and Exponential Smoothing (ES), which are widely used due to their simplicity and interpretability. However, these models often struggle with the complex, non-linear, and high-dimensional data commonly found in industrial systems. To address these challenges, we explore Machine Learning techniques, including Support Vector Machine (SVM) and Artificial Neural Network (ANN). These models offer more flexibility and adaptability, often outperforming traditional statistical methods. Furthermore, we investigate the potential of hybrid models, which combine the strengths of different methods to achieve improved prediction performance. These hybrid models result in more accurate and robust forecasts. Finally, we discuss the potential of newly developed generative models such as Generative Adversarial Network (GAN) for time-series forecasting. This review emphasizes the importance of carefully selecting the appropriate model based on specific industry requirements, data characteristics, and forecasting objectives.

Automatic time series forecasting model design based on pruning

Automatic Time Series Forecasting (TSF) model design which aims to help users to efficiently design forecasting models for the given time series data scenarios, is a novel research topic to be urgently solved. In this paper, we propose AutoTS algorithm trying to utilize the existing design skills and design efficient search methods to effectively solve this problem. In AutoTS, we extract effective design experience from the existing TSF works. We allow the effective combination of design experience from different sources, to create an effective search space containing a variety of TSF models to support different TSF tasks. Considering the huge search space, in AutoTS, we propose a two-stage pruning strategy to reduce the search difficulty and improve the search efficiency. Specifically, at the beginning of the search phase, we apply the vertical pruning method to quickly optimize each module of the TSF model. In the middle of the search, we turn to apply the horizontal pruning method to filter out less effective options for each module according to the learned experience and optimize modules. Moreover, in AutoTS, we introduce the knowledge graph to reveal associations between module options. We make full use of this relational information to learn higher-level features of each module option, to further improve the search quality. Experimental results show that AutoTS is well-suited for the TSF area. On some general datasets, our algorithms discover the assembled model with better performance. Additionally, compared to existing NAS algorithms, they exhibit higher search efficiency and greater search potential than the manually designed ones.

A TCN-Linear Hybrid Model for Chaotic Time Series Forecasting.

The applications of deep learning and artificial intelligence have permeated daily life, with time series prediction emerging as a focal area of research due to its significance in data analysis. The evolution of deep learning methods for time series prediction has progressed from the Convolutional Neural Network (CNN) and the Recurrent Neural Network (RNN) to the recently popularized Transformer network. However, each of these methods has encountered specific issues. Recent studies have questioned the effectiveness of the self-attention mechanism in Transformers for time series prediction, prompting a reevaluation of approaches to LTSF (Long Time Series Forecasting) problems. To circumvent the limitations present in current models, this paper introduces a novel hybrid network, Temporal Convolutional Network-Linear (TCN-Linear), which leverages the temporal prediction capabilities of the Temporal Convolutional Network (TCN) to enhance the capacity of LSTF-Linear. Time series from three classical chaotic systems (Lorenz, Mackey-Glass, and Rossler) and real-world stock data serve as experimental datasets. Numerical simulation results indicate that, compared to classical networks and novel hybrid models, our model achieves the lowest RMSE, MAE, and MSE with the fewest training parameters, and its R2 value is the closest to 1.

Addressing the Non-Stationarity and Complexity of Time Series Data for Long-Term Forecasts

Real-life time series datasets exhibit complications that hinder the study of time series forecasting (TSF). These datasets inherently exhibit non-stationarity as their distributions vary over time. Furthermore, the intricate inter- and intra-series relationships among data points pose challenges for modeling. Many existing TSF models overlook one or both of these issues, resulting in inaccurate forecasts. This study proposes a novel TSF model designed to address the challenges posed by real-life data, delivering accurate forecasts in both multivariate and univariate settings. First, we propose methods termed “weak-stationarizing” and “non-stationarity restoring” to mitigate distributional shift. These methods enable the removal and restoration of non-stationary components from individual data points as needed. Second, we utilize the spectral decomposition of weak-stationary time series to extract informative features for forecasting. To learn features from the spectral decomposition of weak-stationary time series, we exploit a mixer architecture to find inter- and intra-series dependencies from the unraveled representation of the overall time series. To ensure the efficacy of our model, we conduct comparative evaluations against state-of-the-art models using six real-world datasets spanning diverse fields. Across each dataset, our model consistently outperforms or yields comparable results to existing models.

Carbon Emissions and Vegetation Dynamics: Assessing the Spatiotemporal Environmental Impacts of Hydropower Dams in the Lancang River Basin

Recent decades in the Lancang River Basin have witnessed extensive construction of hydropower dams, profoundly impacting the local environment. Utilizing high-precision satellite data, we conducted a comprehensive analysis of vegetation cover and carbon emissions, integrating data-driven time series and spatial analysis models to capture both temporal and spatial dynamics. Our findings reveal that hydropower dam construction in the Lancang River Basin has significantly promoted vegetation restoration and growth, concurrently facilitating a reduction in regional carbon emissions. Employing deep learning models for time-series prediction, we observed a substantial increase in the sum of the local normalized difference vegetation index (NDVI) post-construction, with an average rise of from 16.15% to a maximum of 20.12% during the pivotal hydropower dams’ operational phase. Between 2001 and 2020, the construction of hydropower dams in the basin corresponded to notable changes in ecological and carbon metrics. Specifically, vegetation cover expansion intensity (VCEI) reversed from a negative mean of −0.009 to a positive mean of 0.008. Additionally, the carbon emission intensity (CEI) around these dams drastically reduced, shifting from an average of 0.877 to 0.052. Importantly, the Global Moran’s I for VCEI significantly increased from 0.288 pre-2016 to 0.679 post-2015, reflecting a stronger spatial autocorrelation in vegetation patterns. Accordingly, these findings illustrate the complex interplay between hydropower dams and environmental outcomes, underscoring the critical role of pivotal hydropower dam construction in ecological improvement. The research results have improved and complemented those of previous studies on the environmental impact of hydraulic engineering, providing valuable insights for the construction management and policy formulation of hydropower dams in other similar river basins around the world.

An Investigation into the Pricing and Replenishment Strategy of Vegetable Commodities in Fresh Food Shops

Fresh supermarkets face challenges in marketing and inventory management due to the short shelf life and high loss rate of fresh vegetables. This necessitates the immediate development of a practical replenishment plan and pricing strategy. The paper commences by analyzing the sales distribution patterns of various vegetable categories, identifying leaves and flowers as having notably higher sales compared to other categories. Subsequent examination of the sales trend over time reveals variations in the sales of different categories or individual vegetables across months or quarterly periods, indicating a degree of seasonality. This study calculates the Spearman correlation coefficient for each category and individual product sales, presenting the results in a heat map. Findings indicate a significant positive correlation among all vegetable categories except for solanaceae, implying bundled sales or a consumer tendency to purchase specific vegetable categories concurrently. Consequently, when formulating replenishment and pricing strategies, considering a rational sales mix is crucial. The paper constructs a linear regression model, solving the equation using the least squares approach, and identifies a negative association between sales volume and cost-plus pricing. To accommodate seasonal patterns, the Prophet time series model forecasts wholesale prices for various vegetable categories during July 1-7, 2023. The objective is to maximize profit by accurately predicting vegetable costs. A nonlinear programming model is formulated to achieve this objective, determining the maximum revenue from days 1 to 7 as ￥(108953.65 - K), where K represents the fixed indirect cost. The model optimizes replenishment and pricing strategies for different categories.

LSTM-Based Time Series Prediction and Optimization of Replenishment Pricing Model

Historical sales data play a significant role in the formulation of future replenishment and pricing schemes. In this paper, on the one hand, the correlation relationship between the sales volume of various categories and individual vegetables is investigated by data visualization and calculation of correlation coefficients in different time dimensions, and their "complementary" or "synchronous" relationships are obtained; on the other hand, a more reasonable vegetable replenishment volume and pricing scheme are predicted and obtained through the LSTM time-series prediction model and linear programming model, which are both accurate and accurate at the same time. On the other hand, the LSTM time series prediction model and linear programming model are used to predict and obtain a more reasonable total vegetable replenishment and pricing scheme, which is accurate and flexible at the same time.

Vegetable Automated Pricing Model Based on Particle Swarm Algorithm

Fresh grocery supermarkets face challenges such as the short shelf life of vegetables, quality degradation over time, and difficulty in selling products the following day. Considering these issues, they restock early in the morning without complete knowledge of the specific items or purchase prices, based on historical sales and demand data. To address this issue, this paper first visualizes the data to demonstrate a positive correlation between sales volume and time. Next, to address the issue of predicting the total daily restocking quantity and pricing strategy from July 1 to July 7, 2023, the document utilizes data from the previous three years to establish an ARIMA time series forecast model to predict the costs and sales volume of six vegetables for the first week of July 2023. Finally, to devise the replenishment quantity and pricing strategy for individual items on July 1, 2023, the paper filters the data for all vegetable items from June 24th to 30th, 2023, and combines it with the particle swarm optimization algorithm to obtain the results. This study is pragmatic, providing a feasible solution to the problem of automatic pricing for fresh vegetables.

From simple to complex: a sequential method for enhancing time series forecasting with deep learning

Abstract Time series forecasting is a well-known deep learning application field in which previous data are used to predict the future behavior of the series. Recently, several deep learning approaches have been proposed in which several nonlinear functions are applied to the input to obtain the output. In this paper, we introduce a novel method to improve the performance of deep learning models in time series forecasting. This method divides the model into hierarchies or levels from simpler to more complex ones. Simpler levels handle smoothed versions of the input, whereas the most complex level processes the original time series. This method follows the human learning process where general/simpler tasks are performed first, and afterward, more precise/harder ones are accomplished. Our proposed methodology has been applied to the LSTM architecture, showing remarkable performance in various time series. In addition, a comparison is reported including a standard LSTM and novel methods such as DeepAR, Temporal Fusion Transformer, NBEATS and Echo State Network.