Search

Paired organs like eyes, ears, and lungs in humans exhibit similarities, and data from these organs often display remarkable correlations. Accounting for these correlations could enhance classification models used in predicting disease phenotypes. To our knowledge, there is limited, if any, literature addressing this topic, and existing methods do not exploit such correlations. For example, the conventional approach treats each ear as an independent observation when predicting audiometric phenotypes and is agnostic about the correlation of data from the two ears of the same person. This approach may lead to information loss and reduce the model performance. In response to this gap, particularly in the context of audiometric phenotype prediction, this paper proposes new quadratic discriminant analysis (QDA) algorithms that appropriately deal with the dependence between ears. We propose two‐stage analysis strategies: (1) conducting data transformations to reduce data dimensionality before applying QDA; and (2) developing new QDA algorithms to partially utilize the dependence between phenotypes of two ears. We conducted simulation studies to compare different transformation methods and to assess the performance of different QDA algorithms. The empirical results suggested that the transformation may only be beneficial when the sample size is relatively small. Moreover, our proposed new QDA algorithms performed better than the conventional approach in both person‐level and ear‐level accuracy. As an illustration, we applied them to audiometric data from the Medical University of South Carolina Longitudinal Cohort Study of Age‐related Hearing Loss. In addition, we developed an R package, PairQDA, to implement the proposed algorithms. [ABSTRACT FROM AUTHOR]

There has been a rapid evolution of tree-based ensemble algorithms which have outperformed deep learning in several studies, thus emerging as a competitive solution for many applications. In this study, ten tree-based ensemble algorithms (random forest, bagging meta-estimator, adaptive boosting (AdaBoost), gradient boosting machine (GBM), extreme gradient boosting (XGBoost), light gradient boosting (LightGBM), histogram-based GBM, categorical boosting (CatBoost), natural gradient boosting (NGBoost), and the regularised greedy forest (RGF)) were comparatively evaluated for the enhancement of Copernicus digital elevation model (DEM) in an agricultural landscape. The enhancement methodology combines elevation and terrain parameters alignment, with feature-level fusion into a DEM enhancement workflow. The training dataset is comprised of eight DEM-derived predictor variables, and the target variable (elevation error). In terms of root mean square error (RMSE) reduction, the best enhancements were achieved by GBM, random forest and the regularised greedy forest at the first, second and third implementation sites respectively. The computational time for training LightGBM was nearly five-hundred times faster than NGBoost, and the speed of LightGBM was closely matched by the histogram-based GBM. Our results provide a knowledge base for other researchers to focus their optimisation strategies on the most promising algorithms. [ABSTRACT FROM AUTHOR]

In digital soil mapping, machine learning is gradually replacing traditional statistical models because of their greater flexibility and better prediction performance. However, unlike traditional models, a notable drawback of machine learning models is that they are “black-box” in nature due to their limited ability to provide comprehensive interpretations for their predictions. Explainable machine learning (XML) methods provide visualisations that can be used to aid in understanding predictions made by machine learning models. Popular model-agnostic visualisation methods include partial dependence plots, independent conditional expectation curves, and partial dependence plots produced with Shapley values. These methods require that covariates are uncorrelated which could be restrictive. For cases where covariates are correlated, an alternative approach is the Accumulated Local Effect plot, which however is limited to depicting one or two covariates at a time. Another disadvantage of the above mentioned methods is that no readily available goodness-of-fit metric is available. In this paper we propose the use of a principal component analysis biplot as a model-agnostic method to gain insight into machine learning predictions in digital soil mapping. A biplot is a powerful visualisation tool that is used to seek patterns in multivariate data. A biplot does not require covariates included in the visualisation to be uncorrelated, and furthermore, an analytically derived goodness-of-fit metric is provided which allows the user to evaluate the accuracy of the approximation. We present examples from a case study in South Africa in which soil organic carbon is mapped with a random forest model. Our findings show that biplots can provide meaningful interpretations for predictions, making it a worthy addition to the XML toolkit.

Best linear unbiased prediction (BLUP) is widely used in plant research to address experimental variation. For phenotypic values, BLUP accuracy is largely dependent on properly controlled experimental repetition and how variable components are outlined in the model. Thus, determining BLUP robustness implies the need to evaluate contributions from each repetition. Here, we assessed the robustness of BLUP values for simulated or empirical phenotypic datasets, where the BLUP value and each experimental repetition served as dependent and independent (feature) variables, respectively. Our technique incorporated machine learning and partial dependence. First, we compared the feature importance estimated with the neural networks. Second, we compared estimated average marginal effects of individual repetitions, calculated with a partial dependence analysis. We showed that contributions of experimental repetitions are unequal in a phenotypic dataset, suggesting that the calculated BLUP value is likely to be influenced by some repetitions more than others (such as failing to detect simulated true positive associations). To resolve disproportionate sources, variable components in the BLUP model must be further outlined. [ABSTRACT FROM AUTHOR]

Dams are complex systems that involve both the structure itself and its foundation. Rheological phenomena, expansive reactions, or alterations in the geotechnical parameters of the foundation, among others, result in non-reversible and cumulative modifications in the dam response, leading to trends in the monitoring data series. The accurate identification and definition of these trends to study their evolution are key aspects of dam safety. This manuscript proposes a methodology to identify trends in dam behavioural data series by identifying the influence of the time variable on the predictions provided by the ML models. Initially, ICE curves and SHAP values are employed to extract temporal dependence, and the ICE curves are found to be more precise and efficient in terms of computational cost. The temporal dependencies found are adjusted using a GWO algorithm to different function characteristics of irreversible processes in dams. The function that provides the best fit is selected as the most plausible. The results obtained allow us to conclude that the proposed methodology is capable of obtaining estimates of the most common trends that affect movements in concrete dams with greater precision than the statistical models most commonly used to predict the behaviour of these types of variables. These results are promising for its general application to other types of dam monitoring data series, given the versatility demonstrated for the unsupervised identification of temporal dependencies. [ABSTRACT FROM AUTHOR]

As more machine learning and deep learning models are applied in studying the quantitative relationship between the climate and terrestrial vegetation growth, the uncertainty of these advanced models requires clarification. Partial dependence plots (PDPs) are one of the most widely used methods to estimate the marginal effect of independent variables on the predicted outcome of a machine learning model, and it is regarded as the main basis for conclusions in relevant research. As more controversies regarding the reliability of the results of the PDPs emerge, the uncertainty of the PDPs remains unclear. In this paper, we experiment with real, remote sensing data to systematically analyze the uncertainty of partial dependence relationships between four climate variables (temperature, rainfall, radiation, and windspeed) and vegetation growth, with one conventional linear model and six machine learning models. We tested the uncertainty of the PDP curves across different machine learning models from three aspects: variation, whole linear trends, and the trait of change points. Results show that the PDP of the dominant climate factor (mean air temperature) and vegetation growth parameter (indicated by the normalized difference vegetation index, NDVI) has the smallest relative variation and the whole linear trend of the PDP was comparatively stable across the different models. The mean relative variation of change points across the partial dependence curves of the non-dominant climate factors (i.e., radiation, windspeed, and rainfall) and vegetation growth ranged from 8.96% to 23.8%, respectively, which was much higher than those of the dominant climate factor and vegetation growth. Lastly, the model used for creating the PDP, rather than the relative importance of these climate factors, determines the fluctuation of the PDP output of these climate variables and vegetation growth. These findings have significant implications for using remote sensing data and machine learning models to investigate the quantitative relationships between the climate and terrestrial vegetation. [ABSTRACT FROM AUTHOR]

In digital soil mapping, machine learning is gradually replacing traditional statistical models because of their greater flexibility and better prediction performance. However, unlike traditional models, a notable drawback of machine learning models is that they are "black-box" in nature due to their limited ability to provide comprehensive interpretations for their predictions. Explainable machine learning (XML) methods provide visualisations that can be used to aid in understanding predictions made by machine learning models. Popular model-agnostic visualisation methods include partial dependence plots, independent conditional expectation curves, and partial dependence plots produced with Shapley values. These methods require that covariates are uncorrelated which could be restrictive. For cases where covariates are correlated, an alternative approach is the Accumulated Local Effect plot, which however is limited to depicting one or two covariates at a time. Another disadvantage of the above mentioned methods is that no readily available goodness-of-fit metric is available. In this paper we propose the use of a principal component analysis biplot as a model-agnostic method to gain insight into machine learning predictions in digital soil mapping. A biplot is a powerful visualisation tool that is used to seek patterns in multivariate data. A biplot does not require covariates included in the visualisation to be uncorrelated, and furthermore, an analytically derived goodness-of-fit metric is provided which allows the user to evaluate the accuracy of the approximation. We present examples from a case study in South Africa in which soil organic carbon is mapped with a random forest model. Our findings show that biplots can provide meaningful interpretations for predictions, making it a worthy addition to the XML toolkit. [Display omitted] • Biplots can aid in understanding machine learning predictions with respect to several covariates. • Biplots do not require covariates to be uncorrelated. • Biplots allow for regional and national interpretations of predictions. • Goodness-of-fit statistics are available for biplots. [ABSTRACT FROM AUTHOR]

Investigating the contributions of the factors influencing lake water level and their related changes with regard to hydraulic facilities is vital for understanding the driving mechanism of water level variations under the manifold pressures from anthropogenic activities and climate change. In this study, a random forest (RF) model was used to investigate the changes of the relationship between water level and discharge of the Yangtze River and local tributaries in Poyang Lake, China, based on daily hydrological data from 1980 to 2018. The results indicated that RF exhibited robust capability for water level prediction in Poyang Lake, with average R2 of 0.95, 0.88, 0.92, and 0.94 for the dry, rising, wet, and recession seasons, respectively. Predictor importance analysis showed that the discharge of the Yangtze River had greater influence on the water level than the discharge of local tributaries except for the dry season in Poyang Lake, where the influence on the water level was evident with discharge less than 5,000 m3/s. The influence of the Yangtze River also showed a clear attenuation pattern as the distance from the outlet of the lake increased, where the water level was constantly regulated by the Yangtze River. In addition, the partial dependence plots also indicated that the Yangtze River discharge changes after the TGD operation have resulted in remarkable water level decreases in the wet and recession seasons, especially for the recession period. Meanwhile, a slight increase in water level was predicted under identical discharge of local catchment in the dry season, which was only concentrated in the outlet of the lake. This study indicated the RF model as a robust technique for water level predictions and attribution analysis under multiple temporal and spatial scales. Moreover, this study confirmed the uneven influences of the Yangtze River and local tributaries on water level across different seasons, gauging stations, and phases.

This paper reviews and advocates against the use of permute-and-predict (PaP) methods for interpreting black box functions. Methods such as the variable importance measures proposed for random forests, partial dependence plots, and individual conditional expectation plots remain popular because they are both model-agnostic and depend only on the pre-trained model output, making them computationally efficient and widely available in software. However, numerous studies have found that these tools can produce diagnostics that are highly misleading, particularly when there is strong dependence among features. The purpose of our work here is to (i) review this growing body of literature, (ii) provide further demonstrations of these drawbacks along with a detailed explanation as to why they occur, and (iii) advocate for alternative measures that involve additional modeling. In particular, we describe how breaking dependencies between features in hold-out data places undue emphasis on sparse regions of the feature space by forcing the original model to extrapolate to regions where there is little to no data. We explore these effects across various model setups and find support for previous claims in the literature that PaP metrics can vastly over-emphasize correlated features in both variable importance measures and partial dependence plots. As an alternative, we discuss and recommend more direct approaches that involve measuring the change in model performance after muting the effects of the features under investigation. [ABSTRACT FROM AUTHOR]

We contribute to the methodological literature on the assessment of health inequalities by applying an algorithmic approach to evaluate the capabilities of socioeconomic variables in predicting the prevalence of non-communicable diseases in a Norwegian health survey.We use data from the seventh survey of the population based Tromsø Study (2015–2016), including 11,074 women and 10,009 men aged 40 years and above. We apply the random forest algorithm to predict four non-communicable disease outcomes (heart attack, cancer, diabetes and stroke) based on information on a number of social root causes and health behaviours. We evaluate our results using the classification error, the mean decrease in accuracy, partial dependence statistics.Results suggest that education, household income and occupation to a variable extent contribute to predicting non-communicable disease outcomes. Prediction misclassification ranges between 25.1% and 35.4% depending on the non-communicable diseases under study. Partial dependences reveal mostly expected health gradients, with some examples of complex functional relationships. Out-of-sample model validation shows that predictions translate to new data input. Algorithmic modelling can provide additional empirical detail and metrics for evaluating heterogeneous inequalities in morbidity. The extent to which education, income and occupation contribute to predicting binary non-communicable disease outcomes depends on both non-communicable diseases and socioeconomic indicator. Partial dependences reveal that social gradients in non-communicable disease outcomes vary in shape between combinations of non-communicable disease outcome and socioeconomic status indicator. Misclassification rates highlight the extent of variation within socioeconomic groups, suggesting that future studies may improve predictive accuracy by exploring further subpopulation heterogeneity. [ABSTRACT FROM AUTHOR]

In this paper, some new partial dependent reliability models are obtained by using copula function under the different complex systems. Some explicit expressions of the series and parallel dependent system reliability model are given respectively, such as the complete dependence, complete independence and partial dependence. Then, the different reliability functions of dependent systems are compared, and some specific methods are obtained in order to improve system reliability under the given parameters: (1) Engineers can improve the system reliability by changing the dependency between components, (2) Engineers can improve the system reliability by changing the number of dependent components, (3) Engineers can improve the system reliability by building a new structure dependent system. Finally, some numerical examples are presented to illustrate the results obtained in this article. [ABSTRACT FROM AUTHOR]

Many studies of Mediterranean ecosystems have analyzed vegetation recovery after a wildfire based on fieldwork or remote sensing; however, only a few have adopted a multi-approach assessment. The aim of this study is to determine the viability of a multi-approach using vegetation and remote sensing to observe vegetation recovery time in areas with different wildfire severity. The study area is located in a Mediterranean forest of North-east Spain. After a wildfire, low-, medium- and high-severities with an unburned control were delimited and inventoried at short-, medium- and long-terms using a 20-m transect; measurements were taken in a 1-m width. In each area, vegetal richness (S), diversity (H') and density (D) were measured using fieldwork. The differenced Normalized Burnt Ratio (dNBR) and Mean Decrease Accuracy (%incMSE) were calculated and quantified. Both methods result to be accurate in studying plant density. The dNBR index decreases over time as an effect of the disappearance of fire disturbance. Topographic and vegetation variables help explain the fire severity at very-short and short-terms, while at medium- and long-terms any explanatory power is virtually lost. Partial dependence allowed us to identify those areas that suffered higher fire severity and vegetal evolution over time. [ABSTRACT FROM AUTHOR]

Bayesian confirmation theory is rife with confirmation measures. Many of them differ from each other in important respects. It turns out, though, that all the standard confirmation measures in the literature run counter to the so-called “Reverse Matthew Effect” (“RME” for short). Suppose, to illustrate, that H1 and H2 are equally successful in predicting E in that pE|H1/pE=pE|H2/pE>1. Suppose, further, that initially H1 is less probable than H2 in that p(H1). Then by RME it follows that the degree to which E confirms H1 is greater than the degree to which it confirms H2. But by all the standard confirmation measures in the literature, in contrast, it follows that the degree to which E confirms H1 is less than or equal to the degree to which it confirms H2. It might seem, then, that RME should be rejected as implausible. Festa (Synthese 184:89-100, 2012), however, argues that there are scientific contexts in which RME holds. If Festa’s argument is sound, it follows that there are scientific contexts in which none of the standard confirmation measures in the literature is adequate. Festa’s argument is thus interesting, important, and deserving of careful examination. I consider five distinct respects in which E can be related to H, use them to construct five distinct ways of understanding confirmation measures, which I call “Increase in Probability”, “Partial Dependence”, “Partial Entailment”, “Partial Discrimination”, and “Popper Corroboration”, and argue that each such way runs counter to RME. The result is that it is not at all clear that there is a place in Bayesian confirmation theory for RME. [ABSTRACT FROM AUTHOR]

The deep eutectic solvent pretreatments of lignocellulosic biomass were investigated by machine learning methods. Principal component analysis, partial least squares, linear regression, optimized gradient boosting, artificial neural networks, and random forests were used for reveal mechanisms and inner interaction relationships. The dependences of the pretreatment effect on the variables of the reaction conditions, the DES properties and the lignocellulosic biomass properties were analyzed. The influences of various variables on pretreatment process and partial dependence of lignin removal and carbohydrate recovery were studied. The temperature in the reaction conditions, the hydrophilicity in the DES characteristic parameters, and the hemicellulose content of raw lignocellulose components were the top three most influential factors for the changes in lignin removal, which accounted for 30%, 11%, and 6%, respectively. [Display omitted] • Machine learning prediction of DES pretreatment was studied. • The partial dependences of the pretreatment effect on variables were analyzed. • Random forests revealed mechanisms and inner interaction relationships. • This study could help to guide the reaction design, screen the DES for biom [ABSTRACT FROM AUTHOR]

• Describe a set of methods for the interpretation of complex mapping models. • Methods are not dependent on a specific type of model. • Reveal global model functioning and local drivers of soil variation. • Provide new insights into complex mapping models. • Might assist formulating hypotheses on the mechanisms driving soil variation. Understanding the spatial variation of soil properties is central to many sub-disciplines of soil science. Commonly in soil mapping studies, a soil map is constructed through prediction by a statistical or non-statistical model calibrated with measured values of the soil property and environmental covariates of which maps are available. In recent years, the field has gradually shifted attention towards more complex statistical and algorithmic tools from the field of machine learning. These models are particularly useful for their predictive capabilities and are often more accurate than classical models, but they lack interpretability and their functioning cannot be readily visualized. There is a need to understand how these models can be used for purposes other than making accurate prediction and whether it is possible to extract information on the relationships among variables found by the models. In this paper we describe and evaluate a set of methods for the interpretation of complex models of soil variation. An overview is presented of how model-independent methods can serve the purpose of interpreting and visualizing different aspects of the model. We illustrate the methods with the interpretation of two mapping models in a case study mapping topsoil organic carbon in France. We reveal the importance of each driver of soil variation, their interaction, as well as the functional form of the association between environmental covariate and the soil property. Interpretation is also conducted locally for an area and two spatial locations with distinct land use and climate. We show that in all cases important insights can be obtained, both into the overall model functioning and into the decision made by the model for a prediction at a location. This underpins the importance of going beyond accurate prediction in soil mapping studies. Interpretation of mapping models reveal how the predictions are made and can help us formulating hypotheses on the underlying soil processes and mechanisms driving soil variation. [ABSTRACT FROM AUTHOR]

• Machine learning method is firstly used to systematically study the ES of PCCs. • A database is assembled including 177 flexural-dominant PCCs with 42 features. • A VEL model with feature selection is proposed to predict the ES of PCCs. • The VEL model is experimentally verified and compared with empirical formulas. • The VEL model is interpreted using the combination of PDA and ICE. Predicting effective stiffness (ES) of precast concrete columns (PCCs) is an essential topic when PCCs are applied to structures in seismic zones. However, existing researches provide nearly no method especially for predicting the ES of PCCs, partly due to the high complexity of this issue. This study aims to firstly develop a machine learning (ML) model for predicting the ES of PCCs based on voting ensemble learning (VEL) algorithm which adopted different ML algorithms: support vector regression (SVR), random forest regression (RFR), and gradient boosting tree regression (GBTR) to respectively establish three base models. 177 flexural-dominant PCCs with various construction details were collected from the literature to assemble an experimental database, in which each specimen has 42 features. A ML model establishment proceeding was conducted, involving data preparation, feature selection, and hyperparameter tuning. Experimental verification was conducted to assess the VEL, SVR, RFR, and GBTR models, under the scenarios with/without feature selection. A comparison was conducted on the prediction performance between the existing empirical formulas and the VEL model with feature selection which was further interpreted using the combination of partial dependence analysis (PDA) and individual conditional expectation (ICE). Results show that the VEL algorithm can improve the accuracy and reliability of predicting the ES of PCCs. The VEL model with feature selection is proposed because it eliminates 21 negligible features and still presents far better prediction performance compared with the empirical formulas. The effect of each of the features considered in the proposed VEL model is recognized and depicted. Besides the four parameters considered in the existing formulas, another five parameters are also identified to have non-negligible influence on the ES of PCCs. Despite some limitations such as relatively insufficient number of data points, restricted range of input parameters, and lack of mechanical explanations, the proposed VEL model still firstly provides an accessible way to predict the ES of PCCs and can give inspiration for future researchers. [ABSTRACT FROM AUTHOR]

The chronic partial-dependence diseases are those whereby the individual is capable of facilitating self-care, but needs a caregiver to provide them support, assistance in the disease process, help to contribute to their self-care and thus, fulfill the therapeutic regimen. The labor of the caregiver triggers an interruption in their care, leading to a dysfunction in the quality of life, which should be evaluated to ensure their welfare, taking into account the responsibility when making decisions compared to family care. Caring for a patient with chronic partial-dependence disease implies the experience of changes in their daily lives, related to the transformation of roles, responsibilities, changes in lifestyles and supportive relationship. The purpose of this review article is to provide theoretical foundations that allow nursing professionals to recognize not only the patient with chronic disease requires care, but also his caregiver in order to opportunely identify their needs and develop timely interventions to improve the life quality of both indeed. Therefore, there will be a thorough revision of the classification of chronic diseases according to their dependency, the concept of caregivers of people with chronic partial dependence disease, their quality of life, and the role of professional nursing in the care of these caregivers. [ABSTRACT FROM AUTHOR]

Research Highlights: Forest phenology and productivity were responsive to seasonal heating and chilling accumulation, but responses differed across the temperature range. Background and Objectives: Temperate forests have responded to recent climate change worldwide, but the pattern and magnitude of response have varied, necessitating additional studies at higher spatial and temporal resolutions. We investigated climatic drivers of inter-annual variation in forest phenology and productivity across the Algonquin-to-Adirondacks (A2A) conservation corridor of eastern North America. Methods: We used remotely sensed indices from the AVHRR sensor series and a suite of gridded climate data from the Daymet database spanning from 1989–2014. We used random forest regression to characterize forest–climate relationships between forest growth indices and climatological variables. Results: A large portion of the annual variation in phenology and productivity was explained by climate (pR2 > 80%), with variation largely driven by accumulated heating and chilling degree days. Only very minor relationships with precipitation-related variables were evident. Conclusions: Our results indicate that anthropogenic climate change in the A2A has not yet reached the point of triggering widespread changes in forest phenology and productivity, but the sensitivity of forest growth to inter-annual variation in seasonal temperature accumulation suggests that more temperate forest area will be affected by climate change as warming continues. [ABSTRACT FROM AUTHOR]

The radioactive gas radon (Rn) is considered as an indoor air pollutant due to its detrimental effects on human health. In fact, exposure to Rn belongs to the most important causes for lung cancer after tobacco smoking. The dominant source of indoor Rn is the ground beneath the house. The geogenic Rn potential (GRP) - a function of soil gas Rn concentration and soil gas permeability - quantifies what "earth delivers in terms of Rn" and represents a hazard indicator for elevated indoor Rn concentration. In this study, we aim at developing an improved spatial continuous GRP map based on 4448 field measurements of GRP distributed across Germany. We fitted three different machine learning algorithms, multivariate adaptive regression splines, random forest and support vector machines utilizing 36 candidate predictors. Predictor selection, hyperparameter tuning and performance assessment were conducted using a spatial cross-validation where the data was iteratively left out by spatial blocks of 40 km*40 km. This procedure counteracts the effect of spatial auto-correlation in predictor and response data and minimizes dependence of training and test data. The spatial cross-validated performance statistics revealed that random forest provided the most accurate predictions. The predictors selected as informative reflect geology, climate (temperature, precipitation and soil moisture), soil hydraulic, soil physical (field capacity, coarse fraction) and soil chemical properties (potassium and nitrogen concentration). Model interpretation techniques such as predictor importance as well as partial and spatial dependence plots confirmed the hypothesized dominant effect of geology on GRP, but also revealed significant contributions of the other predictors. Partial and spatial dependence plots gave further valuable insight into the quantitative predictor-response relationship and its spatial distribution. A comparison with a previous version of the German GRP map using 1359 independent test data indicates a significantly better performance of the random forest based map. Unlabelled Image • Mapping of the geogenic radon potential as hazard indicator for indoor radon • Comparison of three machine learning algorithms • Application of spatial cross-validation using spatial blocks to split the data • Partial and spatial dependence plots reveal predictor-response relationship • Random forest GRP map outperforms previous maps using geostatistics [ABSTRACT FROM AUTHOR]