Article
Computer Science, Artificial Intelligence
Ronit Jaiswal, Girish K. Jha, Rajeev Ranjan Kumar, Kapil Choudhary
Summary: The study developed a deep long short-term memory (DLSTM) based model for accurate forecasting of nonstationary and nonlinear agricultural prices series. The DLSTM model, advantageous in capturing nonlinear and volatile patterns, demonstrated superiority in price forecasting ability.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
A. Cabrera, L. G. B. Ruiz, D. Criado-Ramon, C. D. Barranco, M. C. Pegalajar
Summary: This paper presents the implementation and analysis of two approaches (fuzzy and conventional) for energy modeling. The results show that while nonfuzzy models provide more variability and less robustness, the fuzzy solution may be helpful for energy predictions.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2023)
Article
Forestry
Anna Kozuch, Dominika Cywicka, Krzysztof Adamowicz
Summary: The majority of timber price forecasting studies have relied on ARIMA/SARIMA models, while VAR and ETS models have been used less frequently. ANN methodology has not been employed for forecasting timber prices in primary timber markets. This study compares RBF and MLP artificial neural networks with the Prophet procedure and classical models (ARIMA, ETS, BATS, and TBATS) for timber price forecasting in Poland. MLP outperforms other models in terms of price change and level forecasting. ANN models better fit minimum and maximum values compared to classical models. The Prophet procedure yields the lowest quality projections.
Article
Computer Science, Artificial Intelligence
Sourav Kumar Purohit, Sibarama Panigrahi, Prabira Kumar Sethy, Santi Kumari Behera
Summary: Accurate prediction of crop prices is crucial for farmers and government. This study proposes hybrid methods to predict the prices of three commonly used vegetable crops in India, showing superiority over statistical models and machine learning models through extensive statistical analyses.
APPLIED ARTIFICIAL INTELLIGENCE
(2021)
Article
Chemistry, Analytical
Milutin Pavicevic, Tomo Popovic
Summary: As artificial neural network architectures become more efficient in time-series prediction tasks, their application in predicting day-ahead electricity price and demand becomes more attractive. This paper compares different neural network models and demonstrates the promising efficiency of neural networks in short-term electricity prediction, especially when using methods combining fully connected layers and recurrent neural or temporal convolutional layers.
Article
Multidisciplinary Sciences
Vitor Hugo Serravalle Reis Rodrigues, Paulo Roberto de Melo Barros Junior, Euler Bentes dos Santos Marinho, Jose Luis Lima de Jesus Silva
Summary: Developing accurate models for groundwater control is crucial for managing and planning water resources from aquifer reservoirs. The proposed Wavelet Gated Multiformer combines the strengths of a vanilla Transformer and a Wavelet Crossformer to improve the model's predictive capabilities by computing the relationships between time-series points and finding trending periodic patterns. This model outperforms other transformer-like models in terms of Mean Absolute Error reduction.
SCIENTIFIC REPORTS
(2023)
Article
Economics
Malte Lehna, Fabian Scheller, Helmut Herwartz
Summary: The variability of the day-ahead electricity spot price has increased due to the significant increase in renewable energies in electricity production. In this study, four different approaches, including (S)ARIMA(X), LSTM, CNN-LSTM, and VAR models, were compared for forecasting the German day-ahead electricity spot price. The LSTM model achieved the best overall forecasting performance, while the two-stage VAR model performed well for shorter prediction horizons. Moreover, combining both methods resulted in improved electricity spot price forecasts.
Review
Physics, Nuclear
Sichen Li, Andreas Adelmann
Summary: Particle accelerators generate structured data with clear optimization goals and precise control requirements, making them ideal for data-driven research methodologies. The data obtained from sensors and monitors inside the accelerator can be analyzed as multivariate time series. The application of data-driven time series forecasting methods is particularly promising in the field of accelerator control and diagnostics.
PHYSICAL REVIEW ACCELERATORS AND BEAMS
(2023)
Article
Engineering, Civil
Julien Monteil, Anton Dekusar, Claudio Gambella, Yassine Lassoued, Martin Mevissen
Summary: This work investigates the use of deep learning models for long-term large-scale traffic prediction tasks, focusing on scalability. By analyzing 14 weeks of speed observations from over 1000 segments in downtown Los Angeles, different machine learning and deep learning predictors were studied, along with their scalability to larger areas. The study shows that modeling temporal and spatial features into deep learning predictors can be beneficial for long-term predictions, while simpler predictors achieve satisfactory performance for link-based and short-term forecasting, with a trade-off in prediction accuracy, horizon, training time, and model sizing discussed.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Rakshitha Godahewa, Kasun Bandara, Geoffrey Webb, Slawek Smyl, Christoph Bergmeir
Summary: Ensembling techniques are used to improve the performance of Global Forecasting Models (GFM) and univariate models in heterogeneous datasets. A new clustered ensembles methodology is proposed to train multiple GFMs on different clusters of series, achieving higher accuracy than baseline models.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Environmental Studies
Ali Can Ozdemir, Kurtulus Bulus, Kasim Zor
Summary: This study proposes the use of recurrent neural networks, specifically long short-term memory (LSTM) and gated recurrent unit (GRU) networks, based on deep learning algorithms to forecast nickel price variations. The results show that both networks are highly effective and successful, with low prediction error rates. Furthermore, the GRU networks outperform the LSTM networks in terms of computational time.
Article
Entomology
Ramana Narava, Sai Ram D. Kumar, Jagdish Jaba, Anil P. Kumar, Ranga G. Rao, Srinivasa Rao, Suraj Prashad Mishra, Vinod Kukanur
Summary: This research aimed to develop a forecast model using ARIMA and ANN to predict the population dynamics of Helicoverpa armigera. Comparative analysis showed that the ANN based on feed forward neural networks is best suited for effective pest prediction. The developed ARIMA model can help farmers predict the population dynamics of H. armigera and implement timely control measures.
JOURNAL OF INSECT SCIENCE
(2022)
Article
Energy & Fuels
Haolin Yang, Kristen R. Schell
Summary: The study investigates the impact of transfer learning on price prediction DNN representations, showing that it improves accuracy. The GRU-TL architecture, pre-trained on a hybrid dataset, outperforms statistical and deep learning benchmarks. Transfer learning enables the pre-trained DNN representation to learn target dataset features more accurately.
Article
Computer Science, Artificial Intelligence
Andreas Kanavos, Fotios Kounelis, Lazaros Iliadis, Christos Makris
Summary: This paper focuses on the analysis and modeling of passenger demand dynamics in the aviation industry, proposing a method using time series and deep learning techniques to forecast aviation demand, and developing related models. The results of the study show that the proposed methods exhibit satisfactory accuracy and robustness in predicting air travel demand.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Thermodynamics
Ruan Luzia, Lihki Rubio, Carlos E. Velasquez
Summary: Several studies have focused on improving forecasting techniques for capturing multiple patterns in time series. The advancement in computing hardware has made it possible to solve complex equations using large amounts of data, such as neural networks. However, time series methods like ARIMA can also provide good approximations with low computational resources. To enhance ARIMA approximations, they can be combined with techniques like Wavelet Transform or Fourier Transform. This study evaluates the suitability of using artificial neural networks, ARIMA combined with Wavelet Transform, or Fourier Transform to make predictions for different time horizons and frequencies. The results indicate that artificial neural networks perform better for short-term horizons, ARIMA with Fourier Transform provides the best approximation for monthly time series and any time horizon, and ARIMA with Wavelet Transform offers the best approximation for medium-term and long-term periods at any time frequency.