Article
Automation & Control Systems
Lorenzo Nespoli, Vasco Medici
Summary: This paper presents a computationally efficient algorithm for fitting multivariate boosted trees and proves that multivariate trees outperform univariate trees when there is prediction correlation. The algorithm also allows for arbitrary regularization of predictions to enforce properties like smoothness, consistency, and functional relations. Applications and numerical results related to forecasting and control are presented.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Automation & Control Systems
Yichen Zhou, Giles Hooker
Summary: This paper examines a novel gradient boosting framework for regression, which regularizes gradient boosted trees through subsampling and a modified shrinkage algorithm. The resulting algorithm, Boulevard, is shown to converge as the number of trees grows, and a central limit theorem is demonstrated for its limit, providing a characterization of uncertainty for predictions. Simulation study and real world examples support both the predictive accuracy of the model and its limiting behavior.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Computer Science, Interdisciplinary Applications
Thibaut Vaulet, Maya Al-Memar, Hanine Fourie, Shabnam Bobdiwala, Srdjan Saso, Maria Pipi, Catriona Stalder, Phillip Bennett, Dirk Timmerman, Tom Bourne, Bart De Moor
Summary: This study developed a clinical model using machine learning algorithms and an interpretability strategy for predicting first trimester viability. The results showed that gradient boosted algorithms performed similarly to traditional logistic regression models in terms of discrimination and calibration, and were more robust in handling missing values and feature selection.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Article
Computer Science, Artificial Intelligence
Igor Ilic, Berk Gorgulu, Mucahit Cevik, Mustafa Gokce Baydogan
Summary: Time series forecasting involves developing a model based on past observations to predict future events. The explainable boosted linear regression (EBLR) algorithm enhances predictions by explaining errors and incorporating nonlinear features. It provides interpretable results and high predictive accuracy, making it a promising method for time series forecasting.
PATTERN RECOGNITION
(2021)
Article
Demography
Jack Baker, David Swanson, Jeff Tayman
Summary: Small-area population forecasting faces challenges such as small population sizes, shifting population dynamics, data availability, and the evolution of census geographies. Machine learning techniques, specifically boosted regression trees, have shown to have greater accuracy and produce fewer extreme outliers in population forecasts compared to traditional methods.
POPULATION RESEARCH AND POLICY REVIEW
(2023)
Article
Multidisciplinary Sciences
Isabelle Austin-Zimmerman, Daniel F. Levey, Olga Giannakopoulou, Joseph D. Deak, Marco Galimberti, Keyrun Adhikari, Hang Zhou, Spiros Denaxas, Haritz Irizar, Karoline Kuchenbaecker, Andrew Mcquillin, John Concato, Daniel J. Buysse, J. Michael Gaziano, Daniel J. Gottlieb, Renato Polimanti, Murray B. Stein, Elvira Bramon, Joel Gelernter
Summary: This study investigates the genetic basis of sleep duration and identifies 84 independent risk loci for short sleep and 1 locus for long sleep. It also reveals causal associations between sleep and psychiatric traits.
NATURE COMMUNICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Yichen Zhou, Giles Hooker
Summary: This paper studies the use of gradient boosted decision trees as coefficient-deciding functions in varying coefficient models. Compared to traditional methods, boosted trees are more flexible as they do not require structural assumptions. Through empirical research, the proposed method demonstrates advantages in terms of training speed, prediction accuracy, and interpretability.
DATA MINING AND KNOWLEDGE DISCOVERY
(2022)
Article
Automation & Control Systems
Asad Haris, Noah Simon, Ali Shojaie
Summary: In this paper, we propose a unified framework for estimating and analyzing high-dimensional generalized additive models. The framework defines a large class of penalized regression estimators and presents an efficient computational algorithm. We prove min-max optimal convergence bounds for this class and characterize the rate of convergence when a compatibility condition is not met. Additionally, we show the link between the optimal penalty parameters for structure and sparsity penalties in our framework.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Computer Science, Theory & Methods
Estevao B. Prado, Rafael A. Moral, Andrew C. Parnell
Summary: Bayesian additive regression trees (BART) is a successful tree-based machine learning method used for regression and classification problems. This paper introduces an extension of BART called model trees BART (MOTR-BART), which considers piecewise linear functions at node levels for prediction, capturing local linearities more efficiently and requiring fewer trees for equal or better performance compared to BART.
STATISTICS AND COMPUTING
(2021)
Article
Biology
Seonghyun Jeong, Subhashis Ghosal
Summary: In this study, posterior contraction rates in sparse high-dimensional generalized linear models using priors incorporating sparsity were investigated. Results showed that Bayesian methods achieved convergence properties analogous to lasso-type procedures in many generalized linear models, such as logistic regression and Poisson regression.
Article
Psychology, Multidisciplinary
Paolo Montuori, Luigi Mauro Cennamo, Michele Sorrentino, Francesca Pennino, Bartolomeo Ferrante, Alfonso Nardo, Giovanni Mazzei, Sebastiano Grasso, Marco Salomone, Ugo Trama, Maria Triassi, Antonio Nardone
Summary: An incorrect posture can lead to stress on the spine and musculoskeletal disorders. This study aims to assess the determinants of posture in a metropolitan area and found that age and education are the main drivers of correct posture. Despite good knowledge and attitudes towards posture, only a small percentage of the sample consulted specialists for posture issues.
BEHAVIORAL SCIENCES
(2023)
Article
Computer Science, Information Systems
Marcin Czajkowski, Krzysztof Jurczuk, Marek Kretowski
Summary: By integrating the lasso estimator into the tree induction process, the interpretability of the decision tree can be controlled and its overall performance improved.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Interdisciplinary Applications
Ping -Yang Chen, Hsing-Ming Chang, Yu -Ting Chen, Jung-Ying Tzeng, Sheng-Mao Chang
Summary: The TensorTest2D package allows fitting generalized linear models on second-order tensor data, with functions for parameter estimation and hypothesis testing. Two examples are provided to demonstrate its utility in analyzing data from various disciplines, including a tensor regression model for studying multi-omics predictors on drug sensitivity and a logistic tensor regression model for distinguishing handwritten digits. The package also includes a visualization tool for effective classification and variable selection.
Article
Multidisciplinary Sciences
Hui Shan Wong, Md Zobaer Hasan, Omar Sharif, Azizur Rahman
Summary: Since November 2019, countries worldwide have experienced the devastating consequences of the Covid-19 pandemic. This research focused on the external demographic factors and their relation to the spread of Covid-19 in Malaysia. The study found a strong positive correlation between total population and Covid-19 cases, while a weak positive relationship was observed with population density. These findings highlight the importance of considering population size in intervention planning and managing future virus outbreaks in Malaysia.
Article
Statistics & Probability
Jan Pablo Burgard, Patricia Doerr
Summary: Mixed models are commonly used in social and economic analysis. Designing surveys with non-ignorable features can introduce bias in regression parameters. To address this issue, we propose a survey weighted generalized linear mixed model and evaluate its performance through simulation studies and empirical analysis.
JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS
(2022)
Article
Environmental Sciences
Kosuke Nakanishi, Tetsuyuki Ueda, Hiroyuki Yokomizo, Takehiko I. Hayashi
SCIENCE OF THE TOTAL ENVIRONMENT
(2020)
Article
Biology
Joung Hun Lee, Ryo Yamaguchi, Hiroyuki Yokomizo, Mayuko Nakamaru
JOURNAL OF THEORETICAL BIOLOGY
(2020)
Article
Multidisciplinary Sciences
Kosuke Nakanishi, Dai Koide, Hiroyuki Yokomizo, Taku Kadoya, Takehiko Hayashi
SCIENTIFIC REPORTS
(2020)
Article
Environmental Sciences
Kazutaka M. Takeshita, Takehiko Hayashi, Hiroyuki Yokomizo
ENVIRONMENTAL POLLUTION
(2020)
Article
Ecology
Makoto Nishimoto, Tadashi Miyashita, Hiroyuki Yokomizo, Hiroyuki Matsuda, Takeshi Imazu, Hiroo Takahashi, Masami Hasegawa, Keita Fukasawa
Summary: Spatial optimization of capture effort allocation based on past capture records and state-space population models improves control of invasive species, with effectiveness varying depending on total effort level. Spatially heterogeneous density dependence and capture pressure limit snapping turtle abundance, requiring increased effort allocation for successful management.
ECOLOGICAL APPLICATIONS
(2021)
Correction
Multidisciplinary Sciences
Kosuke Nakanishi, Dai Koide, Hiroyuki Yokomizo, Taku Kadoya, Takehiko I. Hayashi
SCIENTIFIC REPORTS
(2020)
Article
Environmental Sciences
Kosuke Nakanishi, Hiroyuki Yokomizo, Takehiko Hayashi
Summary: In recent years, many dragonfly species, including the common Sympetrum frequens in rice paddy fields in Japan, have faced extinction threats. The decline in dragonfly populations was found to be a result of the combined effects of insecticide use and farmland consolidation, highlighting the importance of conservation planning to address habitat degradation and insecticide utilization.
SCIENCE OF THE TOTAL ENVIRONMENT
(2021)
Article
Agricultural Engineering
Kosuke Nakanishi, Nisikawa Usio, Hiroyuki Yokomizo, Tadao Takashima, Takehiko Hayashi
Summary: The study found that the novel insecticide chlorantraniliprole can reduce the emergence rate of dragonfly nymphs into adults, especially for S. infuscatum, but not significantly for S. frequens. This difference could be attributed to differing sensitivity to chlorantraniliprole, varying nymphal stage lengths, or the impact of bottom-up controls on prey organisms sensitive to the insecticide.
PADDY AND WATER ENVIRONMENT
(2022)
Correction
Agricultural Engineering
Kosuke Nakanishi, Nisikawa Usio, Hiroyuki Yokomizo, Tadao Takashima, Takehiko I. Hayashi
PADDY AND WATER ENVIRONMENT
(2022)
Article
Environmental Sciences
Kazutaka M. Takeshita, Takehiko Hayashi, Hiroyuki Yokomizo
Summary: This study provides an overview of the primary data analysis objectives in the early and late chemical management phases, as well as suitable statistical analysis methods for observational datasets. Examples of linear regression analysis procedures using field survey data from Japanese rivers are presented.
INTEGRATED ENVIRONMENTAL ASSESSMENT AND MANAGEMENT
(2022)
Article
Public, Environmental & Occupational Health
Takehiko Hayashi, Ayako Furuhama, Hiroyuki Yokomizo, Hiroshi Yamamoto
Summary: This study quantitatively assessed the efficacy of a derivation procedure for calculating no-effect concentrations for screening assessment of environmental hazards. The results showed that the derivation procedure resulted in high rates of misclassification when only specific data sets were available. The use of additional uncertainty factors improved the consistency of the misclassification rates within the procedure.
Article
Ecology
Minoru Kasada, Yoshihiro Nakashima, Keita Fukasawa, Gota Yajima, Hiroyuki Yokomizo, Tadashi Miyashita
Summary: Understanding the population dynamics of wildlife is crucial for effective management, but accurate estimation of population size is challenging due to limited data availability. By combining camera trap data and administration data, we successfully estimated the population dynamics of wild boar and identified areas where trapping reinforcement is needed for population control.
POPULATION ECOLOGY
(2023)
Article
Environmental Sciences
Kosuke Nakanishi, Hiroyuki Yokomizo, Keiichi Fukaya, Taku Kadoya, Shin-ichiro S. Matsuzaki, Jun Nishihiro, Ayato Kohzu, Takehiko I. Hayashi
Summary: This study used causal impact analysis to evaluate the effects of extreme water-level drawdowns on water quality in Lake Biwa, Japan. The results showed that the timing and magnitude of the extreme drawdowns had different impacts on transparency in different basins of the lake.
SCIENCE OF THE TOTAL ENVIRONMENT
(2022)
Article
Multidisciplinary Sciences
Tetsuro Yoshikawa, Dai Koide, Hiroyuki Yokomizo, Ji Yoon Kim, Taku Kadoya
Summary: Assessing the vulnerability and adaptive capacity of species, communities, and ecosystems is essential for successful conservation. However, climate change introduces extreme uncertainty in assessment pathways, hindering robust decision-making for conservation. In this study, we developed a framework to quantify acceptable uncertainty as a metric of ecosystem robustness, incorporating climate change uncertainty.
SCIENTIFIC REPORTS
(2023)
Article
Environmental Sciences
Kazutaka M. Takeshita, Takehiko Hayashi, Hiroyuki Yokomizo
SCIENCE OF THE TOTAL ENVIRONMENT
(2020)