Model Selection Research Articles

541 Development and External Validation of the “Global Surgical-Site Infection” (GloSSI) Predictive Model in Adult Patients Undergoing Abdominal Surgery

Abstract Aim To develop a novel prognostic score to predict 30-day surgical-site infection (SSI) following gastrointestinal surgery, and externally validate in comparison to existing prognostic models. Method This was a secondary analysis of 3 independent prospective international cohort studies conducted on a global basis. This included adults undergoing gastrointestinal surgery. Model development was performed in the GlobalSurg-2 dataset (January-July 2016). The primary outcome was 30-day SSI, with two predictive techniques explored: penalised regression (LASSO) and machine learning (XGBOOST). Final model selection based on prognostic accuracy (Area Under the Curve [AUC], 95% confidence interval) and clinical utility. Novel and previous scores were externally validated in GlobalSurg-1 (July-November 2014), and GlobalSurg-3 (April-October 2018). Results 30,029 patients were eligible: 14,019 (SSI=12.3%) in the derivation cohort, and 8,464 (SSI=11.4%) and 7,546 (SSI=15.7%) in the validation cohorts (GlobalSurg-1 and GlobalSurg-3 respectively). The LASSO model was selected due to similar discrimination to XGBoost in the GlobalSurg-2 dataset (AUC: 0.738, 0.725-0.750 versus 0.737, 0.709-0.765), but greater explainability. The final GloSSI score included six variables: country income, ASA, diabetes, and operative contamination, approach, and duration. Discrimination remained good on external validation in GlobalSurg-1 (AUC: 0.730, 0.715-0.744), but moderate discrimination in GlobalSurg-3 (AUC: 0.606, 95% CI 0.588-0.623). Nonetheless, this demonstrated superior performance to external validation of all previous models evaluated within GlobalSurg or prior datasets. Conclusions The GloSSI score allowed accurate prediction of SSI with 6 simple variables routinely available at surgery across income settings. This can inform use of intraoperative and postoperative interventions to modify the risk of SSI and minimise associated harm.

Fluvial habitat associations of riverine dragonflies (Odonata, Gomphidae) in the Huron Mountains (Michigan, USA)

AbstractThe presence of sediment‐burrowing clubtail dragonflies (Gomphidae) may be directly related to riverine geomorphic properties. Their synchronous emergence behavior, marked by persistent exuviae, allows for the examination of emerging adult‐stage habitats. Here we explore how emerging adult and benthic larvae are related to hydrogeomorphic factors through regression modeling. Nine sites in two subwatersheds of the Huron Mountain region (Michigan, USA) were surveyed during two periods: May/June 2021 and August 2021. Temperature, pH, dissolved oxygen, conductivity, and discharge were measured concurrent with invertebrate surveys, and field samples of total suspended solids, nutrients, and substrate particle size distribution were also taken. In spring, 317 exuviae were collected, including 153 Gomphidae. In August, 45 larvae were collected including 22 Gomphidae. Spearman's rank correlations preceded robust regression analysis to aid variable selection. Although nonsignificant, Gomphidae exuviae were negatively correlated with conductivity, average water depth, and percent sand while non‐Gomphidae exuviae were positively correlated with the same variables. The model selection found the top models for Gomphidae and non‐Gomphidae exuviae abundance to relate to depth and conductivity, while the top model for Gomphidae larvae was with discharge. All single variable models (discharge, width, and percent sand) had similar AICc criterion values when examining their relationship with non‐Gomphidae larval abundance. Our study demonstrates that larvae of several riverine dragonfly taxa in the Huron Mountains co‐occur despite hydrogeomorphic variation, yet, at emergence, specific taxa may be responding to different stream properties. Understanding the habitat requirements of riverine dragonflies and how these potentially shift throughout development can support conservation efforts.

Frequency-modulated timer regulates torpor–arousal cycles during hibernation in distinct small mammalian hibernators

Hibernation allows mammals to endure harsh seasons by reducing their basal metabolism and body temperature (Tb) to minimize energy expenditure. During hibernation in small animals such as Syrian hamsters and 13-lined ground squirrels, Tb decreases to an ambient level ( < 5 °C) and remains constant for days to weeks in a physiological condition termed deep torpor. Torpor is interrupted by periods of arousal, during which Tb recovers to a euthermic level (approximately 37 °C), and these torpor–arousal cycles are repeated multiple times during hibernation. However, little is known about the mechanisms governing Tb fluctuations during hibernation. In this study, we employed an unbiased model selection approach to Tb data and revealed that a model incorporating frequency modulation quantitatively reproduced Tb fluctuation during hibernation in Syrian hamsters. We found that an unexpectedly long period of 120–430 days modulates a shorter period of several days. In addition, the aforementioned model reproduced Tb fluctuation in 13-lined ground squirrels, which can undergo repeated hibernation according to intrinsic circannual rhythms in constant laboratory conditions. This is the first quantitative study to demonstrate the concerted action of two endogenous periods, one lasting a few days and the other lasting a year, in the torpor–arousal cycles of distinct mammalian hibernators. We anticipate that our theoretical analysis of Tb fluctuation will be a starting point for quantitative comparisons of hibernation patterns across various hibernating species. Furthermore, quantification of Tb data using models will foster our understanding of the molecular mechanisms of hibernation by revealing the biological processes operating within these periods.

Validation of point process predictions with proper scoring rules

AbstractWe introduce a class of proper scoring rules for evaluating spatial point process forecasts based on summary statistics. These scoring rules rely on Monte Carlo approximations of expectations and can therefore easily be evaluated for any point process model that can be simulated. In this regard, they are more flexible than the commonly used logarithmic score and other existing proper scores for point process predictions. The scoring rules allow for evaluating the calibration of a model to specific aspects of a point process, such as its spatial distribution or tendency toward clustering. Using simulations, we analyze the sensitivity of our scoring rules to different aspects of the forecasts and compare it to the logarithmic score. Applications to earthquake occurrences in northern California, United States and the spatial distribution of Pacific silver firs in Findley Lake Reserve in Washington highlight the usefulness of our scores for scientific model selection.

Divergences based Bayesian inference with censored data

. In this work, we deal with some Bayesian inference problems in the presence of right censored data. First, we propose a dual ϕ−divergence Bayes type estimators for parametric models and we establish their asymptotic normality. To establish this result, we need a uniform strong law of large numbers that we prove as well. Then, we consider the problem of prior distributions construction for model selection using ϕ−divergences. Finally, we consider the problem of the predictive density estimation on the basis of ϕ−divergences. We apply an expansion result of the generalized Bayesian predictive density on two parametric models widely used in survival analysis, namely the Weibull and the inverse Weibull model, under right censoring. We also check the performances of our proposed methods through simulations and real data applications. The results of these studies show that our proposed dual ϕ−divergence Bayes type estimators are more robust than other Bayesian estimators. Moreover, the generalized Bayesian predictive density performs better than the classical estimative density especially for the inverse Weibull model.

Integrating genomic data and simulations to evaluate alternative species distribution models and improve predictions of glacial refugia and future responses to climate change

Climate change poses a threat to biodiversity, and it is unclear whether species can adapt to or tolerate new conditions, or migrate to areas with suitable habitats. Reconstructions of range shifts that occurred in response to environmental changes since the last glacial maximum (LGM) from species distribution models (SDMs) can provide useful data to inform conservation efforts. However, different SDM algorithms and climate reconstructions often produce contrasting patterns, and validation methods typically focus on accuracy in recreating current distributions, limiting their relevance for assessing predictions to the past or future. We modeled historically suitable habitat for the threatened North American tree green ash Fraxinus pennsylvanica using 24 SDMs built using two climate models, three calibration regions, and four modeling algorithms. We evaluated the SDMs using contemporary data with spatial block cross‐validation and compared the relative support for alternative models using a novel integrative method based on coupled demographic‐genetic simulations. We simulated genomic datasets using habitat suitability of each of the 24 SDMs in a spatially‐explicit model. Approximate Bayesian computation (ABC) was then used to evaluate the support for alternative SDMs through comparisons to an empirical population genomic dataset. Models had very similar performance when assessed with contemporary occurrences using spatial cross‐validation, but ABC model selection analyses consistently supported SDMs based on the CCSM climate model, an intermediate calibration extent, and the generalized linear modeling algorithm. Finally, we projected the future range of green ash under four climate change scenarios. Future projections using the SDMs selected via ABC suggest only minor shifts in suitable habitat for this species, while some of those that were rejected predicted dramatic changes. Our results highlight the different inferences that may result from the application of alternative distribution modeling algorithms and provide a novel approach for selecting among a set of competing SDMs with independent data.

Beyond Conjugacy for Chain Event Graph Model Selection

Chain event graphs are a family of probabilistic graphical models that generalise Bayesian networks and have been successfully applied to a wide range of domains. Unlike Bayesian networks, these models can encode context-specific conditional independencies as well as asymmetric developments within the evolution of a process. More recently, new model classes belonging to the chain event graph family have been developed for modelling time-to-event data to study the temporal dynamics of a process. However, existing Bayesian model selection algorithms for chain event graphs and its variants rely on all parameters having conjugate priors. This is unrealistic for many real-world applications. In this paper, we propose a mixture modelling approach to model selection in chain event graphs that does not rely on conjugacy. Moreover, we show that this methodology is more amenable to being robustly scaled than the existing model selection algorithms used for this family. We demonstrate our techniques on simulated datasets.

Model selection by pathwise marginal likelihood thresholding

We suggest to estimate a sparse parameter vector in multivariate models through the selection of marginal likelihoods from a potentially large set. The resulting estimator involves an adaptive thresholding mechanism, whereby the marginal estimates are set to zero according to their sequential contribution to the joint information computed along a path of increasingly complex models. The effectiveness of our proposal is illustrated via simulations.

Assessing the impacts of trace element contamination on the physiology and health of seabirds breeding along the western and southern coasts of Portugal

Coastal seabirds serve as sentinels of ecosystem health due to their vulnerability to contamination from human activities. However, our understanding on how contaminant burdens affect the physiological and health condition of seabirds is still scarce, raising the uncertainty on the species' vulnerability vs tolerance to environmental contamination. Here, we quantified 15 Trace Elements (TE) in the blood of gull (yellow-legged gull Larus michahellis and Audouin's gull Ichthyaetus audouinii) and shearwater (Cory's shearwater Calonectris borealis) adults, breeding in five colonies along the Portuguese coastline. Additionally, stable isotopes of carbon (δ13C) and nitrogen (δ15N) were quantified to elucidate foraging habitat and trophic ecology of adults, to identify potential patterns of TE contamination among colonies. We used immuno-haematological parameters as response variables to assess the influence of TE concentrations, stable isotope values, and breeding colony on adults' physiological and health condition. Remarkably, we found blood mercury (Hg) and lead (Pb) concentrations to exceed reported toxicity thresholds in 25% and 13% of individuals, respectively, raising ecotoxicological concerns for these populations. The breeding colony was the primary factor explaining variation in five out of six models, underlining the influence of inherent species needs on immuno-haematological parameters. Model selection indicated a negative relationship between erythrocyte sedimentation rate and both Hg and selenium (Se) concentrations, but a positive relationship with δ13C. The number of immature erythrocyte counts was positively related to Hg and Se, particularly in yellow-legged gulls from one colony, highlighting the colony-site context's influence on haematological parameters. Further research is needed to determine whether essential TE concentrations, particularly copper (Cu) and Se, are falling outside the normal range for seabirds or meet species-specific requirements. Continuous monitoring of non-essential TE concentrations like aluminium (Al), Hg, and Pb, is crucial due to their potential hazardous concentrations, as observed in our study colonies.

A practical guide to implementing artificial intelligence in traditional East Asian medicine research

In this paper, we present a comprehensive guide for implementing artificial intelligence (AI) techniques in traditional East Asian medicine (TEAM) research. We cover essential aspects of the AI model development pipeline, including research objective establishment, data collection and preprocessing, model selection, evaluation, and interpretation. The unique considerations in applying AI to TEAM datasets, such as data scarcity, imbalance, and model interpretability, are discussed. We provide practical tips and recommendations based on best practices and our own experience. The potential of large language models in TEAM research is also highlighted. Finally, we discuss the challenges and future directions of AI application in TEAM, emphasizing the need for standardized data collection and sharing platforms.

Guaranteed-selling service of online used-car reselling platform: Feasibility and applicability compared with conventional resell and agency resell

Information asymmetry in the used-car market causes inconsistent product quality and price perceptions between buyers and sellers. Third-party reselling platforms with expertise can play a positive role in resolving this asymmetry. An emerging guaranteed selling (GS) service is a reselling service developed on online platforms, where the seller is offered a guaranteed selling duration and purchase price before the transaction begins. During this period, sellers can trade directly with buyers. After this period, the platform purchases the unsold used car at the agreed guaranteed purchase price. Although the GS service can reduce sellers’ trading uncertainty and guarantee revenue, it remains unclear whether this emerging service model is feasible. Hence, we verify the feasibility and optimal pricing policy of the GS model. It generates profits by extending the guaranteed selling duration within a threshold limited by the agency rate. Furthermore, we identify the optimal application scenario for the GS model. It is more profitable than the conventional model and agency resell models, considering different consumer preferences. By extending the GS duration and reducing the agency rate within certain limits, the GS model can divert market shares of the conventional resell model and exclude the agency resell model from the platform’s model selection. The interaction between the GS duration and agency rate determines these limits. This study provides valuable insights and managerial guidelines for used-car reselling platforms.

Mapping the Lequesne Functional Index Into the EQ-5D-5L Utility Index in Patients With Knee Osteoarthritis

ObjectiveThis study aimed to map the Lequesne index onto the EuroQol 5 Dimension (EQ-5D-5L) utility index for patients with knee osteoarthritis. MethodsBaseline data from a previous randomized controlled trial were used; 461 patients were involved in the mapping development, and 230 in the validation phase. Various modeling techniques, including generalized linear models, tobit, and beta regression, were used. Factors such as age, sex, and body mass index were considered as covariates. Model selection was based on criteria such as Akaike and Bayesian information criteria, adjusted R2, mean absolute error, and root mean squared error. Validation involved assessing the preselected models using mean absolute error, root mean squared error, and intraclass correlation coefficient. This study follows the Mapping Onto Preference-Based Measures Reporting Standards statement. ResultsFive models were developed, with 2 incorporating age, sex, and/or body mass index along with the Lequesne index showing the best fit across regressions. Validation results were similar for the 3 regressions, with beta regression models exhibiting wider ranges closer to the validation data set. Intraclass correlation coefficient values were better for beta regression models. Both models tended to overpredict for lower EQ-5D-5L values and underpredict for better health status. ConclusionThese mapping functions, the first of their kind, effectively translate the Lequesne index to EQ-5D-5L values in patients with knee osteoarthritis. They demonstrate satisfactory fit and precision, providing valuable tools for clinicians and researchers, particularly in situations where generic preference-based health-related quality of life instruments are inaccessible for utility derivation in cost-effectiveness studies.

NeuroHealth guardian: A novel hybrid approach for precision brain stroke prediction and healthcare analytics

Stroke is a severe illness, that requires early stroke detection and intervention, as this would help prevent the worsening of the condition. The research is done to solve stroke prediction problem, which may be divided into a number of sub-problems such as an individual's predisposition to develop stroke. To attain this objective, a multiturn dataset consisting of various health features, such as age, gender, hypertension, and glucose levels, takes a central role. A multiple approach was put forward concentrating on integrating the machine learning techniques, such as Logistic Regression, Naive Bayes, K-Nearest Neighbors, and Support Vector Machine (SV), together to develop an ensemble machine called Neuro-Health Guardian. The hypothesis "Neuro-Health Guardian Model" integrates these algorithms into one, purported to make stroke prediction more accurate. The topic dives into each instance of preparation of data for analysis, data visualization techniques, selection of the right model, training, testing, ensembling, evaluation, and prediction. The models are validated with error rate accounted from their accuracy, precision, recall, F1 score, and finally confusion matrices for a look. The study's result is showing that the ensemble model that combines the multiple algorithms has the edge over them and this is evidently by the fact that it can predict stroke rises. Additionally, accuracy, precision, recall, and F1 scores are measured in all models and the comparison is done to provide a clear comparison of the models' performance. In short, the article presented the formation of the ongoing stroke prediction that revealed the ensemble model as a good anticipation. Precise stroke predisposition forecasting can assist in early intervention thereby preventing stroke-related deaths, and limiting disability burden by stroke. The conclusions that have come out of this study offer a great action item for the development of predictive models related to stroke prevention and treatment.

A comprehensive survey for generative data augmentation

Generative data augmentation (GDA) has emerged as a promising technique to alleviate data scarcity in machine learning applications. This thesis presents a comprehensive survey and unified framework of the GDA landscape. We first provide an overview of GDA, discussing its motivation, taxonomy, and key distinctions from synthetic data generation. We then systematically analyze the critical aspects of GDA—selection of generative models, techniques to utilize them, data selection methodologies, validation approaches, and diverse applications. Our proposed unified framework categorizes the extensive GDA literature, revealing gaps such as the lack of universal benchmarks. The thesis summarizes promising research directions, including , effective data selection, theoretical development for large-scale models’ application in GDA and establishing a benchmark for GDA. By laying a structured foundation, this thesis aims to nurture more cohesive development and accelerate progress in the vital arena of generative data augmentation.

Double penalized variable selection for high-dimensional partial linear mixed effects models

In this study, we address the selection of both fixed and random effects in partial linear mixed effects models. By combining B-spline and QR decomposition techniques, we propose a double-penalized likelihood procedure for both estimating and selecting these effects. Furthermore, we introduce an orthogonality-based method to estimate the non-parametric component, ensuring that the fixed and random effects are separated without any mutual interference. The asymptotic properties of the resulting estimators are investigated under mild conditions. Simulation studies are conducted to evaluate the finite sample performance of the proposed method. Finally, we demonstrate the practical applicability of our methodology by analyzing a real data.

Maximum pitting corrosion depth prediction of buried pipeline based on theory-guided machine learning

Buried pipelines are crucial for the transportation of oil and natural gas resources. However, pipeline failure accidents have frequently occurred due to corrosion. Therefore, an accurate corrosion depth prediction model is necessary for the reliable supply of energy. In this paper, a theory-guided machine learning (ML) model is developed for maximum pitting corrosion depth prediction, the engineering theory and domain knowledge are integrated into feature space to improve the model interpretability. Firstly, several new feature variables are constructed based on the interactions between independent variables. Then, feature importance of all feature variables is obtained using random forest (RF). A hybrid multi-objective grey wolf optimization (HMOGWO) is proposed to optimize the hyper-parameters of RF model, considering feature number, prediction accuracy, and stability simultaneously. Finally, a comprehensive pitting corrosion dataset is utilized for performance evaluation. The results indicate that the proposed theory-guided model can achieve high prediction accuracy and stability, the optimal feature subset can be determined using multi-objective optimization method simultaneously, which solves the problems of model interpretability and feature selection for traditional ML models with the single-objective optimizer. This study is of great significance to the transportation safety of buried pipelines.

A novel ensemble deep reinforcement learning model for short‐term load forecasting based on Q‐learning dynamic model selection

AbstractShort‐term load forecasting is critical for power system planning and operations, and ensemble forecasting methods for electricity loads have been shown to be effective in obtaining accurate forecasts. However, the weights in ensemble prediction models are usually preset based on the overall performance after training, which prevents the model from adapting in the face of different scenarios, limiting the improvement of prediction performance. In order to improve the accurateness and validity of the ensemble prediction method further, this paper proposes an ensemble deep reinforcement learning approach using Q‐learning dynamic weight assignment to consider local behaviours caused by changes in the external environment. Firstly, the variational mode decomposition is used to reduce the non‐stationarity of the original data by decomposing the load sequence. Then, the recurrent neural network (RNN), long short‐term memory (LSTM), and gated recurrent unit (GRU) are selected as the basic power load predictors. Finally, the optimal weights are ensembled for the three sub‐predictors by the optimal weights generated using the Q‐learning algorithm, and the final results are obtained by combining their respective predictions. The results show that the forecasting capability of the proposed method outperforms all sub‐models and several baseline ensemble forecasting methods.

Non-Native Species in Marine Protected Areas: Global Distribution Patterns

Marine protected areas (MPAs) across various countries have contributed to safeguarding coastal and marine environments. Despite these efforts, marine non-native species (NNS) continue to threaten biodiversity and ecosystems, even within MPAs. Currently, there is a lack of comprehensive studies on the inventories, distribution patterns, and effect factors of NNS within MPAs. Here we show a database containing over 15,000 occurrence records of 2,714 marine NNS across 16,401 national or regional MPAs worldwide. To identify the primary mechanisms driving the occurrence of NNS, we utilize model selection with proxies representing colonization pressure, environmental variables, and MPA characteristics. Among the environmental predictors analyzed, sea surface temperature emerged as the sole factor strongly associated with NNS richness. Higher sea surface temperatures are linked to increased NNS richness, aligning with global marine biodiversity trends. Furthermore, human activities help species overcome geographical barriers and migration constraints. Consequently, this influences the distribution patterns of marine introduced species and associated environmental factors. As global climate change continues to alter sea temperatures, it is crucial to protect marine regions that are increasingly vulnerable to intense human activities and biological invasions.

Population size and structure of beisa oryx and gerenuk in Geralle National Park, southeastern Ethiopia.

The study was conducted to determine the population size of endangered Oryx beisa (Rüppell, 1835), and near-threatened Litocranius walleri (Brooke, 1878) of uncertain global population estimates in Geralle National Park, southeastern Ethiopia. Systematic line transects were established with a transect length range of 2.3 to 6.8 km long (a total of 165.4 km long with a sighting distance of 150 m after truncation). The combination of AIC and chi-square p-values was used as model selection criteria for density/population size estimation in distance sampling software. The lowest AIC, ∆AIC (close to zero), and Chi-square tests (p-value > .05) were selected with adequate model fit. The minimum observation was 67 individuals of beisa oryx in the dry season. The maximum observation was 349 individuals of gerenuk during the wet season. The minimum detection probability of oryx was in the wet season (p â = 76 ± 26), and the minimum detection probability of gerenuk was (p â = 75 ± 1) in both seasons. The two-season pooled density analyzed for studied species indicated (0.85 ± 0.34, 1.24 ± 0.47, beisa oryx/km2), and (3.82 ± 0.6, 4.88 ± 0.7 gerenuk/km2) in dry and wet seasons, respectively. It can be concluded from the results of the study that GNP is home to previously undiscovered healthy populations of the endangered beisa oryx and near-threatened gerenuk. So it is recommended to undergo in-depth population studies, including other species available in the national park and their habitat components, so as to design sound, sustainable conservation measures for the wildlife resources in the area.

Performance evaluation of multiclass classification models for ToN-IoT network device datasets

Internet of things (IoT) technology has empowered tangible objects to establish internet connections, facilitating data exchange with computational capabilities. With significant potential across sectors like healthcare, environmental monitoring, and industrial control, IoT represents a promising technological advancement. This study explores datasets from ToN-IoT’s IoT devices, focusing on multi-class classification, including normal and attack classes, with an additional aim of identifying potential attack sub-classes. Datasets comprise various IoT devices, such as refrigerators, garage doors, global positioning systems (GPS) sensors, motion lights, modbus devices, thermostats, and weather sensors. Comparative analysis is conducted between two prominent multiclass classification models, extreme gradient boosting (XGBoost) and light gradient boosting machine (LightGBM), utilizing accuracy and computational time metrics as evaluation criteria. Research findings highlight that the LightGBM model achieves superior accuracy at 78%, surpassing XGBoost 74.31%. However, XGBoost demonstrates an advantage with a shorter computational time of 1.23 seconds, compared to LightGBM 6.79 seconds. This study not only provides insights into multiclass classification model selection but also underscores the crucial consideration of the trade-off between accuracy and computational efficiency in decision-making. Research contributes to advancing our understanding of IoT security through effective classification methodologies. The findings offer valuable information for researchers and practitioners, emphasizing the nuanced decisions needed when selecting models based on specific priorities like accuracy and computational efficiency.