Article
Environmental Sciences
Dimitrios Effrosynidis, Evangelos Spiliotis, Georgios Sylaios, Avi Arampatzis
Summary: This study compares the performance of different methods in univariate time series forecasting and finds that certain regression methods show more promising results than time series methods. However, time series methods such as ARIMA and Theta still exhibit high accuracy, and the choice of method depends on the specific use case and the trade-off between computational time and performance.
SCIENCE OF THE TOTAL ENVIRONMENT
(2023)
Article
Economics
Christos Katris, Manolis G. Kavussanos
Summary: This paper forecasts the daily Baltic Dry Index using time series and machine learning methods, with a focus on model selection and combining forecasts for improved accuracy. Tests reveal comparable performance between time series and machine learning approaches, with robustness checked through detection of breakpoints in the series.
JOURNAL OF FORECASTING
(2021)
Article
Computer Science, Information Systems
Sandeep Kumar Panda, Sachi Nandan Mohanty
Summary: Accurate demand forecasting is crucial in the food industry to minimize waste and loss caused by improper inventory management. This study compares the impact of various factors on demand using the 'Food Demand Forecasting' dataset and proposes a comparative study of seven regressors to forecast the number of orders. The results demonstrate the potential of deep learning models in forecasting, with LSTM showing superiority over other algorithms.
Article
Engineering, Electrical & Electronic
EunGyeong Kim, M. Shaheer Akhtar, O-Bong Yang
Summary: The current photovoltaic power generation systems face irregularity in distribution. Accurate photovoltaic power forecasting is critical for grid-connected systems under changing environmental circumstances. Time series analysis is crucial for predicting photovoltaic output, especially where past solar radiation data or weather parameters are unavailable. This study utilizes various time-series methods, including deep learning and machine learning algorithms, to forecast photovoltaic power generation output for quick response to equipment and panel defects.
ELECTRIC POWER SYSTEMS RESEARCH
(2023)
Article
Mathematics
Konstantin Chirikhin, Boris Ryabko
Summary: The article proposes forecasting methods based on real-world data compressors that can effectively predict univariate and multivariate data with automatic selection of the best algorithm. Additionally, the use of time-universal codes can reduce computation time without sacrificing accuracy.
Article
Multidisciplinary Sciences
P. Villoria Hernandez, I. Marinas-Collado, A. Garcia Sipols, C. Simon de Blas, M. C. Rodriguez Sanchez
Summary: The study developed HelpResponder, a solution for detecting fire hotspots in hostile environments. By adjusting and modeling variables such as temperature, humidity, and air quality, the best predictive models were identified. The system has been tested in various hostile environments and can assist firefighters in responding quickly to emergencies.
SCIENTIFIC REPORTS
(2023)
Article
Energy & Fuels
Jesus Polo, Nuria Martin-Chivelet, Miguel Alonso-Abella, Carlos Sanz-Saiz, Jose Cuenca, Marina de la Cruz
Summary: This paper studies the power forecasting for building integrated PV (BIPV) systems in a vertical facade using machine learning algorithms based on decision trees. The results show the capabilities of Random Forest and XGBoost algorithms to work as regressors in time series forecasting of BIPV power. The mean absolute error in the deterministic forecast, using the most influencing exogenous variables, was around 40% and close below 30% for the south and east array, respectively.
Article
Engineering, Industrial
Dejan Mircetic, Bahman Rostami-Tabar, Svetlana Nikolicic, Marinko Maslaric
Summary: Demand forecasting is essential for supply chain management. This paper examines the performance of different forecasting approaches and finds that combining forecasts leads to more accurate results. Insights are provided for practitioners, and further research in the area is recommended.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH
(2022)
Article
Engineering, Multidisciplinary
Wanie M. Ridwan, Michelle Sapitang, Awatif Aziz, Khairul Faizal Kushiar, Ali Najah Ahmed, Ahmed El-Shafie
Summary: In this study, various Machine Learning models and methods were used to predict rainfall data in Tasik Kenyir, Terengganu, Malaysia. The results showed that ML models had higher accuracy compared to M2 in different time horizons, with the BDTR model performing the best.
AIN SHAMS ENGINEERING JOURNAL
(2021)
Article
Computer Science, Hardware & Architecture
Zhiqiu Yu, Shuo-Yan Chou, Phan Nguyen Ky Phuc, Tiffany Hui-Kuang Yu
Summary: This research examines the sustainability of Taiwan's power supply chain using system dynamics forecasting. The study finds that Taiwan's electricity demand will continue to rise, but electricity prices do not match industrial demand, leading to potential power shortages.
COMPUTER SYSTEMS SCIENCE AND ENGINEERING
(2022)
Article
Mathematics, Interdisciplinary Applications
Serkan Balli
Summary: The Covid-19 pandemic is the most important health disaster the world has faced in the past eight months, predicting its trend has become a challenge. A study analyzed COVID-19 data and proposed a time series prediction model, estimating the global pandemic will peak at the end of January 2021 with approximately 80 million people cumulatively infected.
CHAOS SOLITONS & FRACTALS
(2021)
Article
Computer Science, Artificial Intelligence
Paolo Mancuso, Veronica Piccialli, Antonio M. Sudoso
Summary: This paper introduces a machine learning approach for forecasting hierarchical time series, using a deep neural network to directly generate accurate and reconciled forecasts, while incorporating explanatory variables to improve forecasting accuracy.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Construction & Building Technology
Wen-jing Niu, Zhong-kai Feng
Summary: Accurate runoff forecasting is crucial for ensuring sustainable utilization and management of water resources. Research indicates that support vector machine, Gaussian process regression, and extreme learning machine outperform artificial neural network and adaptive neural based fuzzy inference system in streamflow prediction, emphasizing the importance of selecting appropriate forecasting models based on reservoir characteristics.
SUSTAINABLE CITIES AND SOCIETY
(2021)
Article
Endocrinology & Metabolism
Xiaomin Fu, Yuhan Wang, Ryan S. S. Cates, Nan Li, Jing Liu, Dianshan Ke, Jinghua Liu, Hongzhou Liu, Shuangtong Yan
Summary: This study used machine learning methods to predict blood glucose levels in type 2 diabetic patients. Various parameters influencing blood glucose were analyzed, and the most effective machine learning algorithm for prediction was determined. The results showed that the XGBoost algorithm had the highest accuracy and can be applied in clinical practice to predict and guide the lifestyle of diabetic patients.
FRONTIERS IN ENDOCRINOLOGY
(2023)
Article
Multidisciplinary Sciences
Mariel Abigail Cruz-Najera, Mayra Guadalupe Trevino-Berrones, Mirna Patricia Ponce-Flores, Jesus David Teran-Villanueva, Jose Antonio Castan-Rocha, Salvador Ibarra-Martinez, Alejandro Santiago, Julio Laria-Menchaca
Summary: This paper addresses the problem of forecasting real-life crime and compares four simple and four machine-learning-based ensemble forecasting methods. It also proposes five forecasting techniques to handle the seasonal component of the time series. The results show that the simple moving average with seasonal removal techniques perform the best for these series.