Article
Ecology
Emily Plant, Rachel King, Jarrod Kath
Summary: This study compared the performance of BART and BRT methods in ecological research, suggesting that BART may be more effective in modeling ecological data, with shorter run times and greater functionality. Ecologists using additive regression approaches may benefit from using BART methods alongside more commonly used BRT methods.
ECOLOGICAL INFORMATICS
(2021)
Article
Computer Science, Information Systems
Shankru Guggari, Vijayakumar Kadappa, V Umadevi, Ajith Abraham
Summary: Decision tree is a widely used non-parametric technique in machine learning, data mining and pattern recognition. This study proposes a novel vertical partitioning technique based on the ideas of music rhythm tree, which shows superior performance in terms of stability and handling of class-imbalanced data.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Forestry
Chunyan Wu, Dongsheng Chen, Xiaomei Sun, Shougong Zhang
Summary: Precise quantification of climate-growth relationships is crucial for scientific forest management. This study investigated the response of tree growth to climate at different altitudes using Larix kaempferi trees. The results showed that tree-ring growth (TRG) was correlated with climate differently in different tree classes at different altitudes. TRG was more sensitive to climate at low altitudes, mainly limited by precipitation, while at high altitudes, the climate-growth relationships were opposite. Dominant trees were found to be the best choice for accurately assessing climate-growth relationships.
JOURNAL OF FORESTRY RESEARCH
(2023)
Article
Engineering, Electrical & Electronic
Yu-Wen Huang, Jicheng An, Han Huang, Xiang Li, Shih-Ta Hsiang, Kai Zhang, Han Gao, Jackie Ma, Olena Chubach
Summary: This paper presents the technical details and experimental results for the VVC block partitioning structure, emphasizing its significant coding gains. The new partitioning structure is designed to be more flexible, allowing different types of tree splits and introduces new methods to optimize coding efficiency.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2021)
Article
Chemistry, Multidisciplinary
Ana Sarcevic, Damir Pintar, Mihaela Vranic, Agneza Krajna
Summary: Global networking and the complexity of computer infrastructures require the use of cutting-edge technologies like data analysis, machine learning, and artificial intelligence to ensure network and information system security. However, in high-risk domains, the deployment of black box intelligent systems is hindered by the lack of transparency, especially as machine learning models become more complex. This research focuses on the use of explainable machine learning to extract knowledge from a specific dataset, comparing the knowledge attained through decision tree rules and the SHAP approach, and providing guidelines for different approaches in specific situations.
APPLIED SCIENCES-BASEL
(2022)
Article
Multidisciplinary Sciences
Robert C. Edgar
Summary: This article introduces the Muscle5 algorithm, which generates an ensemble of diverse high-accuracy alignments to provide confidence estimates in alignments, trees, and other inferences. Experimental results show that using this ensemble can confidently resolve topologies with low bootstrap and reveal that some topologies with high bootstrap are incorrect. This method improves confidence assessment in alignment inference.
NATURE COMMUNICATIONS
(2022)
Article
Agronomy
Mario Lillo-Saavedra, Alberto Espinoza-Salgado, Angel Garcia-Pedrero, Camilo Souto, Eduardo Holzapfel, Consuelo Gonzalo-Martin, Marcelo Somos-Valenzuela, Diego Rivera
Summary: Crop yield forecasting is crucial for farmers' decision-making and planning. However, current methods have limitations, such as limited data collection time. This study presents a methodology using unmanned aerial vehicles and multispectral sensors to predict tomato yield at different stages of crop development, achieving a 9.28% error rate.
Article
Computer Science, Artificial Intelligence
Lev Utkin, Andrei Konstantinov
Summary: This paper proposes ensemble-based modifications to simplify the SHapley Additive exPlanations (SHAP) method for the local explanation of a black-box model. The modifications approximate the SHAP by ensembles with a smaller number of features. Three modifications are proposed, namely ER-SHAP, ERW-SHAP, and ER-SHAP-RF. Numerical experiments demonstrate the effectiveness and local interpretability of these modifications.
Article
Computer Science, Software Engineering
Marina Evers, Lars Linsen
Summary: This study proposes a novel visualization method for multi-dimensional parameter-space partitions, which generates parameter-space partitions by analyzing the similarity space of the ensemble's simulation runs and links them to similarity-space visualizations of the ensemble's simulation runs. With this method, parameter-space partitioning can be visually analyzed and interactively refined.
COMPUTERS & GRAPHICS-UK
(2022)
Article
Chemistry, Analytical
Thi-Thu-Huong Le, Haeyoung Kim, Hyoeun Kang, Howon Kim
Summary: This paper proposes an ML-based IDS method that improves attack detection performance and provides explanations of classification decisions using an ensemble tree approach combined with explainable AI. The method is evaluated using large IoT-based IDS datasets, achieving effective attack detection and supporting cybersecurity experts in optimizing and evaluating their judgments based on explanations.
Article
Forestry
Nina Xiong, Yue Qiao, Huiru Ren, Li Zhang, Rihui Chen, Jia Wang
Summary: Accurate estimation of tree biomass is crucial for monitoring and managing forest resources, understanding regional climate change and material cycles. This study analyzes parameters for additive biomass model systems using smaller sample data and establishes two models based on independent diameter and a combined variable of diameter and tree height. By comparing four different approaches, it is found that both the GMM and logarithmic NSUR methods provide satisfactory goodness of fit and estimation precision, with the GMM method yielding better fitting. The GMM method with the combined variable is suggested for calculation and research of single-tree biomass models with small sample sizes.
Article
Computer Science, Artificial Intelligence
Andrea Apicella, Salvatore Giugliano, Francesco Isgro, Roberto Prevete
Summary: A central issue in eXplainable Artificial Intelligence (XAI) is to provide explanations for the behaviors of non-interpretable machine learning models. This paper proposes an XAI framework that utilizes auto-encoders to extract middle-level input features and generate explanations. Experimental results demonstrate the potential applicability of this method in image classification.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Automation & Control Systems
Neetika Bhandari, Payal Pahwa
Summary: This paper compares different clustering methods on secure perturbed data to identify the method that performs better for analyzing data perturbed using Extended NMF, based on various indexes such as Dunn Index, Silhouette Index, Xie-Beni Index, and Davies-Bouldin Index.
INTELLIGENT AUTOMATION AND SOFT COMPUTING
(2023)
Article
Business, Finance
Mariano Gonzalez-Sanchez, Juan M. Nave Pineda
Summary: This study proposes a methodology for estimating market risk based on the decomposition of series into positive outliers, Gaussian central part, and negative outliers. It uses the negative outliers to estimate the cutoff point and the extreme dependence correlation matrix for measuring portfolio risk. Empirical results on a sample consisting of six assets (Bitcoin, Gold, Brent, Standard&Poor-500, Nasdaq, and Real Estate index) show that this methodology outperforms the Kolmogorov-Smirnov distance in terms of normality and volatility of the tail index, with lower capital consumption. It also improves risk measurement compared to the t-Student copula and allows for estimating extreme dependence and corresponding indexes without the implicit restrictions of elliptic and Archimedean copulas.
INTERNATIONAL REVIEW OF FINANCIAL ANALYSIS
(2023)
Article
Computer Science, Information Systems
Najlaa Maaroof, Antonio Moreno, Aida Valls, Mohammed Jabreel, Pedro Romero-Aroca
Summary: Multi-class classification is a fundamental task in Machine Learning, but complex models are often difficult to interpret. This paper presents a novel method called mcFuzzy-LORE for explaining decisions made by fuzzy-based classifiers. It uses fuzzy decision trees to provide human-readable rules that describe the reasoning behind the model's decision. The method was evaluated on a private dataset and outperformed prior methods in generating counterfactual instances.
Article
Mathematics, Interdisciplinary Applications
Christian Y. Robert
Article
Economics
Michel Denuit, Christian Y. Robert
Summary: This paper examines a peer-to-peer insurance scheme where losses are distributed among participants based on their number, and investigates the asymptotic behavior of individual retention levels, cash-backs, and stop-loss premiums as the number of participants increases. The probability of total loss hitting the stop-loss level is also considered. The results are dependent on the rate of increase in global retention level with the number of participants and the existence of the Esscher transform of losses.
INSURANCE MATHEMATICS & ECONOMICS
(2021)
Article
Economics
Michel Denuit, Christian Y. Robert
Summary: Conditional mean risk sharing is effective for distributing total losses within an insurance pool. This paper develops analytical results for this allocation rule in individual risk models with dependence based on network structures. The Ising model is utilized for modeling occurrences and loss amounts in a decomposable graphical model specific to each participant.
ANNALS OF ACTUARIAL SCIENCE
(2022)
Article
Statistics & Probability
Quang Huy Nguyen, Christian Y. Robert
Summary: This paper investigates the problem of probability estimation for a continuous Gaussian random field on a compact set. It proposes a conditional Monte Carlo type estimator and discusses its asymptotic properties.
JOURNAL OF APPLIED PROBABILITY
(2022)
Article
Statistics & Probability
Michel Denuit, Christian Y. Robert
Summary: The study shows that when the distribution of random variables exhibits log-concavity, the vector becomes larger as the sum increases, known as Efron's monotonicity property. Additionally, under the condition of random variables with density functions and bounded second derivatives, the research explores whether Efron's monotonicity property generalizes to sums involving a large number of terms.
JOURNAL OF MULTIVARIATE ANALYSIS
(2021)
