Article
Economics
Holger Dette, Weichi Wu
Summary: We propose an estimator for the high-dimensional covariance matrix of a locally stationary process with a smoothly varying trend. This estimator is used to derive consistent predictors for nonstationary time series. Unlike existing methods, our predictor does not rely on fitting an autoregressive model nor require a vanishing trend. Simulations and a study on financial indices demonstrate the finite sample properties of our methodology.
JOURNAL OF BUSINESS & ECONOMIC STATISTICS
(2022)
Article
Engineering, Marine
Jae-Hyeon Son, Yooil Kim
Summary: This study developed a computational procedure to predict the structural response of a ship voyaging through irregular seaways while considering relevant uncertainties from a probabilistic perspective. The ship's structural response was represented by linear Volterra series and Laguerre polynomials, with unknown coefficients treated as random variables and probability determined through Bayesian linear regression model. Validation was done using a linear oscillator model and practical application involved analyzing experimental ship data for probabilistic predictions of vertical bending moment time series and estimation of fatigue damage using stochastic time series.
Article
Soil Science
Saham Mirzaei, Ali Darvishi Boloorani, Hossein Ali Bahrami, Seyed Kazem Alavipanah, Alijafar Mousivand, Abdul Mounem Mouazen
Summary: This study aimed to minimize the effect of soil moisture on the accuracy of estimating clay, calcium carbonate, and organic carbon content in semi-arid soils through External parameter orthogonalization (EPO) algorithm. Results showed that EPO correction significantly improved the prediction accuracy of these soil properties, especially when customized for different texture classes. Further research is needed to develop methods to eliminate the effects of external parameters caused by increased salt levels in the soil.
SOIL & TILLAGE RESEARCH
(2022)
Article
Plant Sciences
Kengo Sakurai, Yusuke Toda, Hiromi Kajiya-Kanegae, Yoshihiro Ohmori, Yuji Yamasaki, Hirokazu Takahashi, Hideki Takanashi, Mai Tsuda, Hisashi Tsujimoto, Akito Kaga, Mikio Nakazono, Toru Fujiwara, Hiroyoshi Iwata
Summary: This study evaluated the potential application of temporal multispectral imaging for predicting aboveground biomass (AGB) in soybean. The results showed that the prediction accuracy of AGB genotypic values using multispectral and genomic data was higher than using genomic data alone before the flowering stage. The optimal timing for multispectral imaging may depend on the irrigation levels, as the prediction accuracy varied with time, especially under severe drought conditions.
Article
Computer Science, Interdisciplinary Applications
Predrag Popovic, Milan Gocic, Katarina Petkovic, Slavisa Trajkovic
Summary: This paper introduces a model that can predict evapotranspiration values using artificial neural networks. The model is able to accurately predict the values by incorporating external factors and utilizing a combination of feed-forward and recurrent neural networks.
EARTH SCIENCE INFORMATICS
(2023)
Article
Environmental Sciences
Surbhi Kumari, Sunil Kumar Singh
Summary: China, India, and the USA have the highest energy consumption and CO2 emissions globally. This paper presents a prediction of India's detrimental CO2 emissions for the next 10 years based on various statistical models, machine learning models, and a deep learning-based LSTM model. The results show that the LSTM model is the most accurate for CO2 emission prediction.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2022)
Article
Engineering, Mechanical
Xiao-Ping Zhou, Chang-Qing Li
Summary: This paper presents prospective forecasting of laboratory earthquakes in granite samples. The study proposes a weighted linear least squares and 1st-order polynomial model (WLLS-1OPM) for forecasting the timing of earthquakes and determines the most effective parameters. The results demonstrate the high accuracy of the proposed prospective forecasting model.
TRIBOLOGY INTERNATIONAL
(2022)
Article
Chemistry, Multidisciplinary
Caosen Xu, Jingyuan Li, Bing Feng, Baoli Lu
Summary: This paper proposes a model based on multiplexed attention mechanisms and linear transformers to predict financial time series. The results show that the proposed method can effectively improve the prediction accuracy of the model, increase the inference speed of the model, and reduce the number of operations, which has new implications for the prediction of financial time series.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Information Systems
Yang Wang, Lixin Han
Summary: The study addresses the limitations of existing research on recommendation systems by constructing time sequences predictive models and proposing a Hybrid Network Adaptive Time Series recommendation framework to better analyze and utilize hidden structural and temporal information for improved recommendation performance.
INFORMATION PROCESSING & MANAGEMENT
(2021)
Article
Computer Science, Artificial Intelligence
Wonkeun Jo, Dongil Kim
Summary: Deep neural networks are important in machine learning for their excellent prediction performance and versatility. However, they lack explanatory power due to being black-box models. This study proposes a new neural network architecture that includes interpretability for multivariate time-series data. Experimental results show that the interpretable neural architecture performs well in predicting MTS data and provides reasonable importance for each input value.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Mathematics, Applied
Yoshito Hirata, Masanori Shiro
Summary: This article proposes a method to predict the maxima of a flow more accurately by using local cross sections or plates. It provides a theoretical underpinning for observability using local cross sections and improves short-term prediction by employing a generalized prediction error. The approach is demonstrated using rainfall data, where heavier rains may cause casualties.
Article
Computer Science, Artificial Intelligence
Xinyu Chen, Lijun Sun
Summary: This paper proposes a Bayesian temporal factorization (BTF) framework for modeling multidimensional time series, particularly spatiotemporal data, in the presence of missing values. By integrating low-rank matrix/tensor factorization and vector autoregressive (VAR) process, this framework can characterize both global and local consistencies in large-scale time series data.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Yitong Li, Kai Wu, Jing Liu
Summary: This paper proposes a robust time series prediction framework called spARIMA, which reduces noise interference by designing a sequential training scheme in batches based on the degree of noise. spARIMA relies on the differential prediction model in ARIMA and absorbs the advantages of the gradual training scheme in self-paced learning (SPL) to effectively address the instability caused by noise. Furthermore, spARIMA introduces diversity selection to avoid selecting similar samples, using a weighted local complexity-similarity distance expression to represent the diversity of noisy data. Comparative tests with existing ARIMA models on two gradient descent algorithms show that spARIMA not only works well with noisy data, but also performs efficiently with normal data, indicating its generalization ability.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Mathematics
Kun Zhang, Xing Huo, Kun Shao
Summary: Utilizing a temperature time-series prediction model can accurately sense changes in temperature levels in advance, which is important. This study proposes a model that combines the STL decomposition method, the JUST algorithm, and the Bi-LSTM network, achieving accurate temperature predictions in China.
Article
Statistics & Probability
Shuhao Jiao, Alexander Aue, Hernando Ombao
Summary: This article introduces a significant method, called partial functional prediction (PFP), in functional time series analysis for reliable predictions of future functions. The PFP method utilizes both completely observed trajectories and partial information for prediction, with an automatic selection criterion for tuning parameters based on minimizing prediction error.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2023)
Article
Chemistry, Analytical
Yaoli Wang, Yujun Duan, Wenxia Di, Qing Chang, Lipo Wang
Article
Computer Science, Information Systems
Yaoli Wang, Lipo Wang, Fangjun Yang, Wenxia Di, Qing Chang
Summary: The study introduces an improved Elman neural network called Elman-DIOCs with direct input-to-output connections, demonstrating its effectiveness in stock forecasting. Experimental results show that DIOCs lead to significantly better prediction accuracy while requiring fewer hidden neurons. The study argues that DIOCs can improve accuracy and reduce complexity in neural networks for regression or classification tasks with linear components.
INFORMATION SCIENCES
(2021)