Common mistakes in ecological niche models

International Journal of Geographical Information Science, Journal Year: 2020, Volume and Issue: 35(2), P. 213 - 226

Published: July 27, 2020

Ecological niche models (ENMs) are widely used statistical methods to estimate various types of species niches. After lecturing several editions introductory courses on ENMs and reviewing numerous manuscripts this subject, we frequently faced some recurrent mistakes: 1) presence-background modelling methods, such as Maxent or ENFA, if they were pseudo-absence methods; 2) spatial autocorrelation is confused with clustering records; 3) environmental variables a higher resolution than 4) correlations between not taken into account; 5) machine-learning replicated; 6) topographical calculated from unprojected coordinate systems, and; 7) downscaled by resampling. Some these mistakes correspond student misunderstandings corrected before publication. However, other errors can be found in published papers. We explain here why approaches erroneous propose ways improve them.

Language: Английский

---

Top ten hazards to avoid when modeling species distributions: a didactic guide of assumptions, problems, and recommendations

Ecography, Journal Year: 2024, Volume and Issue: 2024(4)

Published: Jan. 31, 2024

Species distribution models, also known as ecological niche models or habitat suitability have become commonplace for addressing fundamental and applied biodiversity questions. Although the field has progressed rapidly regarding theory implementation, key assumptions are still frequently violated recommendations inadvertently overlooked. This leads to poor being published used in real‐world applications. In a structured, didactic treatment, we summarize what our view constitute ten most problematic issues, hazards, negatively affecting implementation of correlative approaches species modeling (specifically those that model by comparing environments species' occurrence records with background pseudoabsence sample). For each hazard, state relevant assumptions, detail problems arise when violating them, convey straightforward existing recommendations. We discuss five major outstanding questions active current research. hope this contribution will promote more rigorous these valuable stimulate further advancements.

Language: Английский

---

How Interpretable Machine Learning Can Benefit Process Understanding in the Geosciences

Earth s Future, Journal Year: 2024, Volume and Issue: 12(7)

Published: July 1, 2024

Abstract Interpretable Machine Learning (IML) has rapidly advanced in recent years, offering new opportunities to improve our understanding of the complex Earth system. IML goes beyond conventional machine learning by not only making predictions but also seeking elucidate reasoning behind those predictions. The combination predictive power and enhanced transparency makes a promising approach for uncovering relationships data that may be overlooked traditional analysis. Despite its potential, broader implications field have yet fully appreciated. Meanwhile, rapid proliferation IML, still early stages, been accompanied instances careless application. In response these challenges, this paper focuses on how can effectively appropriately aid geoscientists advancing process understanding—areas are often underexplored more technical discussions IML. Specifically, we identify pragmatic application scenarios typical geoscientific studies, such as quantifying specific contexts, generating hypotheses about potential mechanisms, evaluating process‐based models. Moreover, present general practical workflow using address research questions. particular, several critical common pitfalls use lead misleading conclusions, propose corresponding good practices. Our goal is facilitate broader, careful thoughtful integration into science research, positioning it valuable tool capable enhancing current

Language: Английский

---

Malware Detection: A Framework for Reverse Engineered Android Applications Through Machine Learning Algorithms

IEEE Access, Journal Year: 2022, Volume and Issue: 10, P. 89031 - 89050

Published: Jan. 1, 2022

Today, Android is one of the most used operating systems in smartphone technology. This main reason, has become favorite target for hackers and attackers. Malicious codes are being embedded applications such a sophisticated manner that detecting identifying an application as malware toughest job security providers. In terms ingenuity cognition, progressed to point where they're more impervious conventional detection techniques. Approaches based on machine learning have emerged much effective way tackle intricacy originality developing threats. They function by first current patterns activity then using this information distinguish between identified threats unidentified with unknown behavior. research paper uses Reverse Engineered applications' features Machine Learning algorithms find vulnerabilities present Smartphone applications. Our contribution twofold. Firstly, we propose model incorporates innovative static feature sets largest datasets samples than methods. Secondly, ensemble i.e., AdaBoost, Support Vector (SVM), etc. improve our model's performance. experimental results findings exhibit 96.24% accuracy detect extracted from applications, 0.3 False Positive Rate (FPR). The proposed ignored detrimental permissions, intents, Application Programming Interface (API) calls, so on, trained feeding solitary arbitrary feature, reverse engineering input machine.

Language: Английский

---

Machine learning of large‐scale spatial distributions of wild turkeys with high‐dimensional environmental data

Ecology and Evolution, Journal Year: 2019, Volume and Issue: 9(10), P. 5938 - 5949

Published: April 24, 2019

Species distribution modeling often involves high-dimensional environmental data. Large amounts of data and multicollinearity among covariates impose challenges to statistical models in variable selection for reliable inferences the effects factors on spatial species. Few studies have evaluated compared performance multiple machine learning (ML) handling multicollinearity. Here, we assessed effectiveness removal correlated regularization cope with ML habitat suitability. Three algorithms maximum entropy (MaxEnt), random forests (RFs), support vector machines (SVMs) were applied original (OD) 27 landscape variables, reduced (RD) 14 highly being removed, 15 principal components (PC) OD accounting 90% variability. The three was measured area under curve continuous Boyce index. We collected 663 nonduplicated presence locations Eastern wild turkeys (Meleagris gallopavo silvestris) across state Mississippi, United States. Of total locations, 453 separated by a distance ≥2 km used train OD, RD, PC data, respectively. remaining 210 validate trained measure performance. had excellent RD MaxEnt SVMs good indicating adequacy default setting Weak RFs through bagging appeared alleviate resulted Regularization may help exploratory suitability wildlife.

Language: Английский

---

Spatial machine learning: new opportunities for regional science

The Annals of Regional Science, Journal Year: 2021, Volume and Issue: 68(3), P. 713 - 755

Published: Dec. 24, 2021

Abstract This paper is a methodological guide to using machine learning in the spatial context. It provides an overview of existing toolbox proposed literature: unsupervised learning, which deals with clustering data, and supervised displaces classical econometrics. shows potential this developing methodology, as well its pitfalls. catalogues comments on usage methods (for locations values, both separately jointly) for mapping, bootstrapping, cross-validation, GWR modelling density indicators. details models, are combined data integration, modelling, model fine-tuning predictions deal autocorrelation big data. The delineates “already available” “forthcoming” gives inspiration transplanting modern quantitative from other thematic areas research regional science.

Language: Английский

---

Predicting accounting fraud using imbalanced ensemble learning classifiers – evidence from China

Accounting and Finance, Journal Year: 2023, Volume and Issue: 63(3), P. 3455 - 3486

Published: Jan. 9, 2023

Abstract The current research aims to launch effective accounting fraud detection models using imbalanced ensemble learning algorithms for China A‐Share listed firms. Based on a sample of 33,544 Chinese firm‐year instances from 1998 2017, this respectively established one logistic regression and four classifiers (AdaBoost, XGBoost, CUSBoost, RUSBoost) by 12 financial ratios 28 raw data. Additionally, we divided the into train test observations evaluate classifiers' out‐of‐sample performance. In detail, applied two metrics, namely, Area under ROC (receiver operating characteristic) curve (AUC) Precision‐Recall (AUPR), discriminability. supplement test, study put forward an algebraic fused model basis introduced sliding window technique. empirical results showed that can detect A‐listed firms far more effectively than model. Moreover, (CUSBoost performed better common (AdaBoost XGBoost) in average. also obtained highest average AUC AUPR among all employed algorithms. Our offer firm support potential role Machine Learning (ML)‐based Artificial Intelligence (AI) approaches reliably predicting with high accuracy. Similarly, settings, our ML‐based AI offers utmost advantage forecasting fraud. Finally, paper fills gap applications

Language: Английский

---

Predicting Habitat Suitability and Conserving Juniperus spp. Habitat Using SVM and Maximum Entropy Machine Learning Techniques

Water, Journal Year: 2019, Volume and Issue: 11(10), P. 2049 - 2049

Published: Sept. 30, 2019

Support vector machine (SVM) and maximum entropy (MaxEnt) learning techniques are well suited to model the habitat suitability of species. In this study, SVM MaxEnt models were developed predict Juniperus spp. in Southern Zagros Mountains Iran. recent decades, drought extension climate alteration have led extensive changes geographical occurrence species its growth regeneration extremely limited area. This study evaluated through spatial modeling predicts appropriate regions for future cultivation resource conservation. We modeled natural an area 700 ha Sepidan Area Fars province using (1) data regarding presence (295 samples) collected field surveys GPS, (2) soil information indices derived from 60 samples area, (3) climatic topographic datasets various sources. total, 15 conditioning factors used approach. Receiver operator characteristic (ROC) curves applied estimate accuracy produced by techniques. Results indicated logical similar under curve (AUC)-ROC values (0.735) (0.728) models. Both methods revealed a significant relationship between distribution factors. Environmental played vital role evaluating sp. as Max Min temperatures annual mean rainfall three most important Finally, with high very landscape conservation was suggested based on model.

Language: Английский

---

Species distribution modeling to inform transboundary species conservation and management under climate change: promise and pitfalls

Frontiers of Biogeography, Journal Year: 2022, Volume and Issue: 14(1)

Published: Jan. 20, 2022

Spatially explicit biogeographic models are among the most used methods in conservation biogeography, with correlative species distribution (SDMs) being popular them. SDMs can identify potential for species’ and community range shifts under climate change, thus inspire, inform, guide complex adaptive management planning efforts such as collaborative transboundary frameworks. However, rarely developed collaboratively, which would be ideal applications of models. Further, that applied to often do not follow best practices field, particularly important change contexts model extrapolation into potentially novel climates is necessary. Thus, while there substantial promise, machine-learning based SDM approaches, also many pitfalls consider when applying conservation, especially context change. Here, we summarize these key steps mitigate them maximize promise facilitate We argue modeling capacity must elevated practitioners they easily implement using SDMs, regarding: 1) avoiding overcomplexity, 2) addressing input data bias, 3) accounting uncertainty extrapolations projections. While our discussion centers mainly on opportunities algorithm, Maxent, suggestions generalized a other tools. Overall, improved training in, tools for, implementation hold great help complex, collaborations long-term

Language: Английский

---

Assessing the applicability of binary land-cover variables to species distribution models across multiple grains

Landscape Ecology, Journal Year: 2024, Volume and Issue: 39(3)

Published: March 4, 2024

Abstract Context Species distribution models are widely used in ecology. The selection of environmental variables is a critical step SDMs, nowadays compounded by the increasing availability data. Objectives To evaluate interaction between grain size and binary (presence or absence water) proportional (proportion water within cell) representation cover variable when modeling bird species distribution. Methods eBird occurrence data with an average number records 880,270 per across North American continent were for analysis. Models (via Random Forest) fitted 57 species, two seasons (breeding vs. non-breeding), at four grains (1 km 2 to 2500 ) using as variable. Results models’ performances not affected type adopted (proportional binary) but significant decrease was observed importance form. This especially pronounced coarser during breeding season. Binary useful finer sizes (i.e., 1 ). Conclusions At more detailed ), simple presence certain land-cover can be realistic descriptor occurrence. particularly advantageous collecting habitat field simply recording significantly less time-consuming than its total area. For grains, we recommend variables.

Language: Английский

---