14 results
Search Results
2. Chaotic time series prediction based on multi-scale attention in a multi-agent environment.
- Author
-
Miao, Hua, Zhu, Wei, Dan, Yuanhong, and Yu, Nanxiang
- Subjects
- *
TIME series analysis , *MULTIAGENT systems , *FORECASTING , *DYNAMICAL systems , *MULTISCALE modeling , *CHAOS theory - Abstract
A new problem at the intersection of multi-agent systems, chaotic time series prediction, and flow map learning is formulated in this paper. The problem involves agents collaborating to track moving targets in chaotic dynamic systems by communicating. Inspired by the multi-scale hierarchical time-stepper (HiTS), a novel Distributed Prediction Network based on Multi-scale Attention (DPNMA) is proposed to fuse predictions from agents at different scales through an enhanced self-attention mechanism. The experimental evaluation demonstrates that DPNMA effectively mitigates cumulative errors and enhances the accuracy and robustness of the predictions, which has important implications for the scenarios where the agents have heterogeneous and constrained capabilities. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
3. A Haavelmo grey model based on economic growth and its application to energy industry investments.
- Author
-
Li, Hui, Nie, Weige, and Duan, Huiming
- Subjects
- *
ENERGY consumption , *GREENHOUSE gases , *ENERGY industries , *ECONOMIC expansion , *ECONOMIC models , *MATHEMATICAL transformations , *ECONOMIC forecasting , *FORECASTING - Abstract
The energy industry is a major source of greenhouse gas emissions, and energy investment is an important regulatory tool to encourage the energy industry to actively respond to climate change and achieve low-carbon development. Therefore, it is of great practical significance to correctly understand the important role of the energy industry, to predict energy investments objectively and accurately, to achieve scientific and rational investment, and make policy recommendations for the energy production and consumption revolution. In this paper, the Haavelmo model of economic growth is introduced into the energy system, using the characteristics of the continuous form of the model to establish the differential equations for the dynamics of fixed asset investment in the energy industry, and Haavelmo's grey prediction model using the grey difference information principle. Meanwhile, the Python program is used to solve the parameters of the new model, and the mathematical transformation is used to find the time response equation of the new model, and the modeling steps and the modeling flow chart of the model are obtained. Finally, the new model will be applied to two types of energy investments in China: total energy industry investment and investment in electricity, steam, hot water production, and supply industry. Both types of energy use the same modeling object and forecast object, and six cases are compared with three grey forecasting models from different perspectives, and their results show that they are much better than the other three grey forecasting models, demonstrating the effectiveness of the new model to effectively forecast energy investments and improve the efficiency of energy industry investments, cultivate healthy and environmentally friendly energy consumption habits. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
4. Complexity-aided time series modeling and forecasting under a decomposition-aggregation framework.
- Author
-
Song, Mingli and Wang, Ruobing
- Subjects
- *
TIME series analysis , *TIME complexity , *HILBERT-Huang transform , *FORECASTING - Abstract
• A novel comprehensive time series prediction framework comprising a decomposition-aggregation process and outputting information granules is proposed. • Complexity is effectively used to guide modeling from two aspects. • The application of GNNs is further extended to time series prediction field. • Granular neural networks are used to capture critical patterns of the original time series. Complexity of time series has always been of significant interest to researchers; however, it is not yet effectively explored in assisting time series prediction modeling. In this study, we develop a decomposition-aggregation time series prediction framework sufficiently applying the complexity concept from two aspects. One aspect is realized during the process of allocating information granularity onto each decomposed subsequence (with comparably lower complexity). The other aspect is realized through constituting the objective function of a prediction model (a granular neural network in this paper) which is built with a subsequence individually. To capture several critical patterns of a time series, it is decomposed into several subsequences and used to train a granular neural network independently. An effective aggregation strategy is then adopted to aggregate those predicted information granules which are the outputs of local granular neural networks. The quality of the global output information granule is assessed by analyzing the relationship of coverage and specificity and keeping in mind the overall information granularity. Experimental studies show that the decomposition-aggregation strategy prediction results are better than the non-decomposed one and the complexity-aided objective function performs better than the objective function without complexity. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
5. Linear asymmetric Laplace fuzzy information granule and its application in short-to-medium term prediction for financial time series.
- Author
-
Yang, Hong and Wang, Lina
- Subjects
- *
TIME series analysis , *LAPLACE distribution , *FUZZY numbers , *FORECASTING , *TIME perspective - Abstract
Gaussian fuzzy information granule (GFIG) and its linear form provide a novel perspective for time series modeling. However, numerous studies show that asymmetric Laplace distribution has unique advantages over the Gaussian one, which motivate us to construct asymmetric Laplace distribution based fuzzy information granules and pursue their properties and applications. Three main contributions are made in this paper. First, asymmetric Laplace fuzzy information granule (ALFIG) on one-dimensional fuzzy number space is proposed, then its linear operations and distance metric are discussed. Second, membership-weighted kernel line is proposed to construct linear-ALFIG for extracting plenty of trend information contained within time series specially. Third, a linear-ALFIG based Long Short-Term Memory model (A-LSTM) is proposed for short-to-medium term prediction of financial time series. Experiment results show that: (i) Fitting errors of linear-ALFIG are significantly lower than that of linear-GFIG for datasets that exactly obey asymmetric Laplace distribution; (ii) A-LSTM has statistically absolute advantages in short-to-medium term prediction under significance level α = 0.1 (in most cases α = 0.05 in fact), which not only predicts direction, amplitude and changepoints of future trends, but also delivers comprehensive, transparent and user-oriented results. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
6. A cluster prediction strategy with the induced mutation for dynamic multi-objective optimization.
- Author
-
Xu, Kangyu, Xia, Yizhang, Zou, Juan, Hou, Zhanglu, Yang, Shengxiang, Hu, Yaru, and Liu, Yuan
- Subjects
- *
EVOLUTIONARY algorithms , *FORECASTING , *MEMETICS - Abstract
Dynamic multi-objective optimization problems (DMOPs) are multi-objective optimization problems in which at least one objective and/or related parameter vary over time. The challenge of solving DMOPs is to efficiently and accurately track the true Pareto-optimal set when the environment undergoes changes. However, many existing prediction-based methods overlook the distinct individual movement directions and the available information in the objective space, leading to biased predictions and misleading the subsequent search process. To address this issue, this paper proposes a prediction method called IMDMOEA, which relies on cluster center points and induced mutation. Specifically, employing linear prediction methods based on cluster center points in the decision space enables the algorithm to rapidly capture the population's evolutionary direction and distributional shape. Additionally, to enhance the algorithm's adaptability to significant environmental changes, the induced mutation strategy corrects the population's evolutionary direction by selecting promising individuals for mutation based on the predicted result of the Pareto front in the objective space. These two complementary strategies enable the algorithm to respond faster and more effectively to environmental changes. Finally, the proposed algorithm is evaluated using the JY, dMOP, FDA, and F test suites. The experimental results demonstrate that IMDMOEA competes favorably with other state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
7. Progressive neural network for multi-horizon time series forecasting.
- Author
-
Lin, Yang
- Subjects
- *
ENERGY consumption forecasting , *TIME series analysis , *FORECASTING , *AUTOREGRESSIVE models , *ENERGY consumption , *PERFORMANCES - Abstract
In this paper, we introduce ProNet, an novel deep learning approach designed for multi-horizon time series forecasting, adaptively blending autoregressive (AR) and non-autoregressive (NAR) strategies. Our method involves dividing the forecasting horizon into segments, predicting the most crucial steps in each segment non-autoregressively, and the remaining steps autoregressively. The segmentation process relies on latent variables, which effectively capture the significance of individual time steps through variational inference. In comparison to AR models, ProNet showcases remarkable advantages, requiring fewer AR iterations, resulting in faster prediction speed, and mitigating error accumulation. On the other hand, when compared to NAR models, ProNet takes into account the interdependency of predictions in the output space, leading to improved forecasting accuracy. Our comprehensive evaluation, encompassing four large datasets, and an ablation study, demonstrate the effectiveness of ProNet, highlighting its superior performance in terms of accuracy and prediction speed, outperforming state-of-the-art AR and NAR forecasting models. • Forecasting of electricity energy consumption and solar energy generation. • Informer-based model marrying forecasting horizon segmentation and variational inference techniques. • Comparing SARIMAX, DeepAR, DeepSSM, LogTrans, N-BEATS, and Informer models forecasting ability. • Performance evaluation based on four real-world datasets. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
8. Quasi-synchronization of parameter mismatch drive-response systems: A self-triggered impulsive control strategy.
- Author
-
Zheng, Huannan, Zhu, Wei, and Li, Xiaodi
- Subjects
- *
FORECASTING - Abstract
The objective of this paper is to research the quasi-synchronization of parameter mismatch drive–response systems via a self-triggered impulsive control (STIC) strategy. In the proposed STIC strategy, the impulsive instants are determined by a carefully crafted self-triggered mechanism (STM). In contrast to the event-triggered impulsive control (ETIC), which continuously supervises an event-triggered condition to determine whether to generate an impulse, the proposed STIC strategy predicts the next impulsive instant using the last impulsive instant and measurable information, eliminating continuous event monitoring. By means of the presented STIC strategy, some sufficient quasi-synchronization conditions and error bound for parameter mismatch drive–response systems are obtained. Moreover, the Zeno-behavior can be eliminated for the impulsive instants. The effectiveness of the proposed STIC strategy is demonstrated through two numerical examples. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
9. Deep graph tensor learning for temporal link prediction.
- Author
-
Liu, Zhen, Li, Zhongyi, Li, Wen, and Duan, Lixin
- Subjects
- *
LOGICAL prediction , *REPRESENTATIONS of graphs , *FORECASTING - Abstract
Knowledge discovery on dynamic graphs has received much attention in recent years. As a key task of dynamic graph research, the goal of temporal link prediction is to accurately predict the time-varying links in dynamic networks. Uncertainty in link emergence is a major challenge in this research, as it is not easy to learn stable and reliable link-level feature representations, which are usually readily available on static graphs. In order to adapt to the ever-changing graph structure, this paper proposes to construct a deep graph tensor learning model, which can capture the contextual characteristics of graph evolution from both the graph structure (spatial) mode and the link sequence (temporal) mode. Therefore, compared to link prediction on static graphs, temporal link prediction can benefit more from the link-level embedding representations coupled with spatio-temporal features. The experimental results on seven public dynamic graph datasets show that the prediction accuracy obtained by the new model is overall better than competing models such as GC-LSTM, EvolveGCN, and HTGN. In the meantime, as a result of getting rid of the traditional RNN learning paradigm, the new model is also significantly better than the traditional temporal graph learning model in terms of training efficiency. • A graph tensor learning-based embedding method for temporal graphs is proposed. • The efficiency of dynamic graph representation learning can be greatly improved by using the graph compact technique. • State-of-the-art accuracy for temporal link prediction can be achieved by the proposed model. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
10. Novel deterministic and probabilistic forecasting methods for crude oil price employing optimized deep learning, statistical and hybrid models.
- Author
-
Purohit, Sourav Kumar and Panigrahi, Sibarama
- Subjects
- *
PETROLEUM sales & prices , *LOGICAL prediction , *MACHINE learning , *DEEP learning , *STATISTICAL models , *FORECASTING - Abstract
• Optimized deep learning, statistical and hybrid models are proposed for deterministic (point) and probabilistic (interval) forecasting of crude oil price (COP). • A DE-DL method employing differential evolution algorithm is proposed to optimize the architecture and other hyper-parameters of deep learning models for a problem. • Recurrence plot is drawn, recurrence quantification analysis is carried out and uncertainty analysis of prediction error is performed using Gaussian and t-Location Scale distributions. • Extensive non-parametric statistical analysis on the obtained results are carried out to rank the optimized models and draw decisive conclusions relating to deterministic and probabilistic forecasting of COP. • Eight research questions related to deterministic and probabilistic forecasting of COP are addressed. In this paper, individual and hybrid methods are proposed employing optimized statistical and deep learning (DL) models for deterministic (point) and probabilistic (interval) forecasting of crude oil price time series. The statistical models are optimized using the Forecast package of R. To enhance the performance of DL models, a novel pruning DE-DL method is proposed, which employs the differential evolution (DE) algorithm to optimize architecture and continuous and discrete-valued hyper-parameters. The proposed DE-DL method is so generic that it can be applied to optimize different DL models for any supervised learning problem. Five DL models (LSTM, BiLSTM, GRU, CNN, and ConvLSTM) are optimized for forecasting monthly crude oil prices and hybridized with an optimized ARIMA model for developing optimized additive and multiplicative hybrid forecasting models. The effectiveness of the proposed methods is evaluated through deterministic and probabilistic forecasting measures, comparing the results with six optimized statistical models, thirteen machine learning models, five optimized DL models, and ten optimized hybrid models. It is observed from the simulation results that the proposed optimized Additive-ARIMA-GRU hybrid model provides statistically superior forecasts, and the t Location Scale distribution is more suitable than the Gaussian distribution for computing reliable prediction intervals with different significance levels. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
11. TRNN: An efficient time-series recurrent neural network for stock price prediction.
- Author
-
Lu, Minrong and Xu, Xuerong
- Subjects
- *
RECURRENT neural networks , *DATA compression , *ARTIFICIAL neural networks , *BACK propagation , *FINANCIAL markets , *FORECASTING - Abstract
Prediction results in big data analysis can vary greatly depending on the data preprocessing methods used. Time series-based processing methods are particularly advantageous for prediction. While popular neural network models such as Back Propagation (BP), Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM) are based on weight, loss function, and other factors, their training efficiency is still relatively low. In this paper, we propose an efficient Time-series Recurrent Neural Network (TRNN) for stock price prediction. In the proposed model, trading volume is established and sliding windows are used to process the time series data. The trends and turning points of the data are extracted according to financial market features, and data compression is achieved. To improve the impact of recent trading volume on the current stock price, the price-volume relationship is upgraded from one dimension to two dimensions based on RNN. The information about trading volume is processed and compressed to establish the TRNN model, which guarantees both accuracy and efficiency. We compare our TRNN model with the original RNN and LSTM models in terms of efficiency and accuracy. We further discuss the feasibility of related expanded schemes of our TRNN model, as well as the extendability of the time-series compression and TRNN model to other fields. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
12. Grey–Markov prediction model based on time-continuous Markov model and Levenberg–Marquardt algorithm.
- Author
-
Zhang, Lei, Li, Ruijiang, and Kang, Shugui
- Subjects
- *
SYSTEMS theory , *PREDICTION models , *LEAST squares , *MODEL theory , *GAUSSIAN distribution , *QUANTILE regression , *FORECASTING - Abstract
• Transition probabilities of traditional Grey–Markov model are obtained by transition frequency with small sample, and the information of small sample is not fully utilized, which lead to unreliable predicted results, we solve this problem by improving Grey–Markov model with time-continuous Markov model and Levenberg–Marquardt algorithm. • It is the first approach to use higher order iterative method to obtain intensities in Grey–Markov model predicting. • Normal distribution quantile of transition probability is used to obtain modified predicting value, and the conclusion that divided number of Markov error states does not affect the modified result is proven. Improved Grey–Markov model in this paper has better interpretability. Grey–Markov model, which combines the advantages of grey system theory and Markov model, has a wide range of applications in various fields. However, reliable predicting can hardly be given for the reason that transition probabilities of traditional Grey–Markov model are obtained by transition frequency with small sample size and the information of small sample size is not fully utilized. To solve this problem, we improve Grey–Markov model based on time-continuous Markov model. First, based on the residual errors of GM (1, 1), we construct the least error square objective function with Kolmogorov forward equations. Second, we improve Levenberg–Marquardt algorithm to obtain the optimal transition intensities and calculate corresponding transition probabilities of Markov model. Third, we obtain predicted value with normal distribution quantile. Not only the improved prediction model has better interpretability, but case study is applied to verify its validity by comparing it with traditional Grey–Markov model. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
13. Depth asynchronous time delay reservoir for nonlinear time series forecasting task.
- Author
-
You, Meiming, Wang, Guoqiang, Yang, Zhao, and Yang, Xuesong
- Subjects
- *
TIME series analysis , *COMPUTER network traffic , *FORECASTING , *LONG-term memory , *NONLINEAR systems , *SHORT-term memory , *INDUSTRIAL research - Abstract
• The model proposed integrates the advantages of each of the width model (DDR) and depth model (DTDR). • A time delay operator is inserted between adjacent layers to control the transmission process of feature information. • The short-term memory capacity of the model proposed was significantly improved. • The model proposed has higher forecasting accuracy for long-term memory datasets. The forecasting of nonlinear time series has a wide range of applications both in theoretical research and industrial production, as the fitting of nonlinear systems, the early warning of sunspot activity and the prediction of network traffic peaks in advance. These tasks require models with strong nonlinear mapping capabilities and sufficient short-term memory capacity. Single-node time delay Reservoir (TDR) have been frequently applied on nonlinear time series forecasting tasks. To address the lack of memory capacity of traditional TDR and its extended version (Delay Decoupled Reservoir (DDR), Deep Time-Delay Reservoir (DTDR)), we propose a Depth Asynchronous Time-Delay Reservoir (DATDR) model. Firstly, the model retains the deep network structure of the DTDR to ensure the dynamic properties of the model, and the history states of each layer are from the previous layers instead of the current layer. Secondly, the same signal input method as the DDR is used, i.e., each layer is fed with the original signal, and this data input method ensures that the feature of signal does not decay gradually during the transmission between layers. Finally, a time delay operator is inserted into two adjacent layers, and the memory capacity of the model could be controlled effectively by adjusting the delay time of the time delay operator to accommodate some time series prediction tasks that require long memory capacity. In this paper, eight datasets from three different types are used to validate the proposed model, including NARMA10-30, Mackey-Glass Chaos time series, He ́non map series, Sunspot Sequence, and two real network traffic datasets. Experiments show that proposed model gets significant improvement in memory capacity compared with TDR, DDR, and DTDR, and performs better for some nonlinear time series datasets especially some long-term memory datasets. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
14. Build interval-valued time series forecasting model with interval cognitive map trained by principle of justifiable granularity.
- Author
-
Ouyang, Chenxi, Yu, Fusheng, Hao, Yadong, Tang, Yuqing, and Jiang, Yanan
- Subjects
- *
TIME series analysis , *COGNITIVE training , *FORECASTING , *PROBLEM solving - Abstract
• Propose a method for constructing interval cognitive maps (ICMs); • Build an interval-valued time series (ITS) forecasting model based on the proposed ICM; • The proposed ICM based ITS forecasting model has higher accuracy; • The proposed ICM based ITS forecasting model can avoid counterintuitive outputs. In the study of time series forecasting based on fuzzy cognitive maps (FCMs), the causalities between past values and future values are represented by real-valued weights in [ - 1 , 1 ]. However, for interval-valued time series (ITS), the causalities are affected by various uncertainties including ways of measuring and ways of intervals influencing intervals and thus involve uncertainty. Therefore, real-valued weights are no longer enough for characterizing such causalities, equipping FCMs with interval-valued weights becomes necessary and resulting in interval cognitive maps (ICMs). In this case, how to determine the interval-valued weights of an ICM becomes a crucial problem. To solve this problem, this paper first proposes the principle of justifiable granularity for interval-valued data, which is guaranteed to accumulate enough experimental evidence and effectively express the ITS, then develops a reasonable method that can optimally determine the interval-valued weights and enable the interval-valued weights having clear semantics. By means of the proposed method for determining interval-valued weights, an ICM-based ITS forecasting model is established, which can not only deal with the uncertainty of causalities between interval-valued data, but also avoid counterintuitive outputs which often appeared in existing ITS forecasting models. Experimental results show the good performance of the proposed forecasting model. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF
Discovery Service for Jio Institute Digital Library
For full access to our library's resources, please sign in.