Partial Dependence Plots Research Articles

Deciphering house prices by integrating street perceptions with a machine-learning algorithm: A case study of Xi'an, China

A comprehensive understanding of house prices and their factors provide insights into the demand for housing while helping policymakers implement measures to manage the housing market. Traditional studies either focus more on linear relationships and ignore complex, non-linear influences or consider neighborhood amenities but lose sight of the streetscape. This study aims to enrich the literature by integrating street-perception characteristics with an interpretable machine-learning technique for modeling house prices. Specifically, street-view images were semantically segmented to quantify street-perception characteristics from five perspectives: greenness, openness, enclosure, walkability, and imageability. By combining the determinants of community attributes and living convenience, 17 explanatory variables were fed into a gradient-boosting decision tree (GBDT) model to estimate housing prices. The results reveal that the model significantly outperforms the linear model (R2 increased by 47.87 %). Additionally, an improvement of 26.15 % (R2) was observed when street-perception characteristics were incorporated. Moreover, complicated non-linear relationships and interaction effects are discussed by visualizing partial dependence plots (PDPs). These findings offer nuanced guidance for improving the neighborhood environment to promote urban equity and develop a sustainable housing market.

Combining SWAT with Machine Learning to Identify Primary Controlling Factors and Their Impacts on Non-Point Source Pollution

Non-point source (NPS) pollution has a complex formation mechanism, and identifying its primary controlling factors is crucial for effective pollution treatment. In this study, the Baixi Reservoir Watershed, characterized by low-intensity development, was selected as the study area. A new methodology combining the Soil and Water Assessment Tool (SWAT) with the Random Forest (RF) algorithm was proposed to comprehensively identify the primary controlling factors of NPS pollution and analyze the interaction between factors. The results of the validated SWAT model showed that the annual intensity of total nitrogen (TN) load range was 0.677–11.014 kg ha−1 yr−1, and the total phosphorus (TP) load per unit area range was 0.020–0.110 kg ha−1 yr−1. Loads of sediment, TP, and TN exhibited significant seasonal variations, particularly in the Baixi basin, where sediment yield had the highest absolute change rate, with a value of up to 232.26. Random Forest models for TN and TP displayed high accuracy (R2 > 0.99) and robust generalization ability. Fertilization, sediment yield, and terrain slope were identified through RF models as the primary factors affecting TN and TP. By graphing partial dependency plots (PDPs) based on the results of the RF models to analyze the interaction between factors, the findings suggest a strong synergistic effect of two combined factors: fertilization and sediment yield. When fertilizer application exceeds 15 kg ha−1 yr−1 and sediment yield exceeds 3 kg ha−1 yr−1, there is a sharp increase in nitrogen and phosphorus load. Through the identification and analysis of the primary controlling factors of NPS pollution, this study provides a solid scientific foundation for developing effective watershed management strategies.

Machine Learning-Based Prediction for High Health Care Utilizers by Using a Multi-Institutional Diabetes Registry: Model Training and Evaluation.

The cost of health care in many countries is increasing rapidly. There is a growing interest in using machine learning for predicting high health care utilizers for population health initiatives. Previous studies have focused on individuals who contribute to the highest financial burden. However, this group is small and represents a limited opportunity for long-term cost reduction. We developed a collection of models that predict future health care utilization at various thresholds. We utilized data from a multi-institutional diabetes database from the year 2019 to develop binary classification models. These models predict health care utilization in the subsequent year across 6 different outcomes: patients having a length of stay of ≥7, ≥14, and ≥30 days and emergency department attendance of ≥3, ≥5, and ≥10 visits. To address class imbalance, random and synthetic minority oversampling techniques were employed. The models were then applied to unseen data from 2020 and 2021 to predict health care utilization in the following year. A portfolio of performance metrics, with priority on area under the receiver operating characteristic curve, sensitivity, and positive predictive value, was used for comparison. Explainability analyses were conducted on the best performing models. When trained with random oversampling, 4 models, that is, logistic regression, multivariate adaptive regression splines, boosted trees, and multilayer perceptron consistently achieved high area under the receiver operating characteristic curve (>0.80) and sensitivity (>0.60) across training-validation and test data sets. Correcting for class imbalance proved critical for model performance. Important predictors for all outcomes included age, number of emergency department visits in the present year, chronic kidney disease stage, inpatient bed days in the present year, and mean hemoglobin A1c levels. Explainability analyses using partial dependence plots demonstrated that for the best performing models, the learned patterns were consistent with real-world knowledge, thereby supporting the validity of the models. We successfully developed machine learning models capable of predicting high service level utilization with strong performance and valid explainability. These models can be integrated into wider diabetes-related population health initiatives.

Determining the Influence of Real Estate Features on Prices with Partial Dependence Plots: A Case Study in Szczecin, Poland

Abstract The study explores the application of Partial Dependence Plots (PDP) in the analysis of real estate features. The study centers on a selected real estate market in Szczecin, Poland, aiming to highlight the efficacy of PDP in understanding and interpreting the complex relationships between various features and property prices. The primary objective is to showcase the potential of PDP in capturing the nuanced interactions between real estate attributes and their impact on market prices. The CatBoost model, known for its robust handling of categorical features and strong predictive capabilities, is employed as the machine learning algorithm for this analysis. The performance of this model will be compared against a traditional multiple linear regression model, providing insights into the advantages of leveraging advanced machine learning techniques in real estate analysis. Results obtained from the analysis will be presented and discussed, shedding light on the interpretability and accuracy of the CatBoost model compared to the traditional linear regression approach. The presentation will conclude with implications for real estate practitioners and researchers, emphasizing the potential for PDP to enhance the transparency and understanding of complex models in the real estate domain. This research contributes to the growing body of knowledge on the application of advanced machine learning techniques in real estate analysis.

Predicting photosynthetic bacteria-derived protein synthesis from wastewater using machine learning and causal inference

Causal inference-assisted machine learning was used to predict photosynthetic bacterial (PSB) protein production capacity and identify key factors. The extreme gradient boosting algorithm effectively predicted protein content, while the gradient boosting decision tree algorithm excelled in predicting protein production, protein productivity, and protein energy yields. Driving factors were identified, with suitable ranges: protein content (pH 6.0–7.5, hydraulic retention time (HRT) < 3.8 d), protein production (biomass > 1.7 g, organic loading rate (OLR) > 9.2 gL–1d–1, temperature 26.7–35.0 °C), protein productivity (HRT < 3.5 d, biomass > 1.6 g, OLR > 10.0 gL–1d–1), and protein energy yields (light energy 0.1–4.4 kWh, biomass 1.7–65.0 g, chemical oxygen demand (COD) 0.1–2.5 gL–1). Illuminance, dissolved oxygen, COD, and COD/total nitrogen ratio were causal factors influencing protein production. Two-dimensional partial dependence plot revealed the interaction between two driving factors. This study enhances information on PSB protein production and offers insights for wastewater treatment and sustainable resource development.

Estimation of compressive strength of concrete with manufactured sand and natural sand using interpretable artificial intelligence

To satisfy the design strength of manufactured sand concrete (MSC) in practical engineering applications, a plethora of geotechnical tests are frequently conducted. An effective approach is imperative to reduce the consumption of labor and resources during these tests. The objective of this paper is to introduce an interpretable machine learning (ML) method to evaluate the compressive strength (CS) of MSC. Firstly, a dataset was established by compiling experimental results from 208 literatures. 3382 data points were selected from the dataset for algorithm training. Recursive Feature Elimination with Cross-Validation (RFECV) was employed to select input parameters. Four algorithms with 12 selected input variables and 1 output variable were constructed to predict CS of MSC using Random Forest (RF), Gradient Boosting Decision Trees (GBDT), eXtreme Gradient Boosting algorithm (XGBoost), and Categorical Boosting (CatBoost). The results show that XGBoost has the highest accuracy and generalization ability (R2=0.934, MAE=3.44, RMSE=5.16, MAPE=0.07). To enhance model transparency, SHapley Additive exPlanations (SHAP) was adopted to explain the underlying predictive mechanisms of ML models. Analyses show that, 1) the cement content, curing time, and water content were the main feature parameters influencing the CS of MSC, 2) the cement content, curing time, and water content have a linear increase, logarithmic increase, and exponential decrease with the CS of MSC, respectively. Partial Dependence Plots (PDP) and Individual Conditional Expectation (ICE) plots were used to further analyze the impacts of these significant influencing factors on the CS of MSC. Additionally, the Local Interpretable Model-Agnostic Explanations (LIME) method was employed to investigate thresholds for various material dosages in MSC containing 5–10 % stone powder. Two typical scenarios were selected for analysis, yielding recommended dosage ranges for concrete of two distinct strengths. Finally, a graphical user interface (GUI) for the CS of MSC has been designed, which might be of great use to material engineers. This provides reference and guidance for concrete engineering practice.

Exploring machine learning to study and predict the chloride threshold level for carbon steel reinforcement

Chloride-induced corrosion of steel reinforcing bar (rebar) is the primary cause of deterioration in reinforced concrete structures, posing a significant infrastructure challenge. The chloride threshold level (CTL) of rebar, which represents the critical amount of chloride needed to initiate active corrosion, is crucial in corrosion and service life prediction models. However, substantial uncertainties and a multitude of influencing factors, along with the absence of a universally accepted testing framework, hinder the achievement of a consistent CTL range for service life models and complicate comparisons of published values. This study addresses these challenges by developing multiple machine learning models to predict CTL, considering 21 carefully selected features. A comprehensive database of 423 data points was compiled from an exhaustive literature review. Seven machine learning models—linear regression, decision tree, random forest, K-nearest neighbors, support vector machine, artificial neural network, and an ensemble model—were developed and optimized. The ensemble model achieved superior prediction performance, with a mean absolute error of 0.218 % by weight of binder, root mean square error of 0.321 %, and a coefficient of determination of 0.751 on unseen CTL data. Partial dependence plots generated using the support vector machine model quantified the effect of each feature on CTL. The random forest model identified SiO₂ binder content and exposed rebar area to chlorides as the most influential factors. The study also examined the impact of supplementary cementitious materials (SCMs), finding that only blast furnace slag positively affected CTL.

Impact of Enterprise Supply Chain Digitalization on Cost of Debt: A Four-Flows Perspective Analysis Using Explainable Machine Learning Methodology

Rising costs, complex supply chain management, and stringent regulations have created significant financial burdens on business sustainability, calling for new and rapid strategies to help enterprises transform. Supply chain digitalization (SCD) has emerged as a promising approach in the context of digitalization and globalization, with the potential to reduce an enterprise’s debt costs. Developing a strategic framework for SCD that effectively reduces the cost of debt (CoD) has become a key academic challenge, critical for ensuring business sustainability. To this end, under the perspective of four flows, SCD is deconstructed into four distinct features: logistics flow digitalization (LFD), product flow digitalization (PFD), information flow digitalization (IFD), and capital flow digitalization (CFD). To precisely measure the four SCD features and the dependent variable, COD, publicly available data from Chinese listed manufacturing enterprises such as annual report texts and financial statement data are collected, and various data mining technologies are also used to conduct data measurement and data processing. To comprehensively investigate the impact pattern of SCD on CoD, we employed the explainable machine learning methodology for data analysis. This methodology involved in-depth data discussions, cross-validation utilizing a series of machine learning models, and the utilization of Shapley additive explanations (SHAP) to explain the results generated by the models. To conduct sensitivity analysis, permutation feature importance (PFI) and partial dependence plots (PDPs) were also incorporated as supplementary explanatory methods, providing additional insights into the model’s explainability. Through the aforementioned research processes, the following findings are obtained: SCD can play a role in reducing CoD, but the effects of different SCD features are not exactly the same. Among the four SCD features, LFD, PFD, and IFD have the potential to significantly reduce CoD, with PFD having the most substantial impact, followed by LFD and IFD. In contrast, CFD has a relatively weak impact, and its role is challenging to discern. These findings provide significant guidance for enterprises in furthering their digitalization and supply chain development, helping them optimize SCD strategies more accurately to reduce CoD.

A machine learning algorithm to explore the drivers of carbon emissions in Chinese cities

As the world’s largest energy consumer and carbon emitter, the task of carbon emission reduction is imminent. In order to realize the dual-carbon goal at an early date, it is necessary to study the key factors affecting China’s carbon emissions and their non-linear relationships. This paper compares the performance of six machine learning algorithms to that of traditional econometric models in predicting carbon emissions in China from 2011 to 2020 using panel data from 254 cities in China. Specifically, it analyzes the comparative importance of domestic economic, external economic, and policy uncertainty factors as well as the nonparametric relationship between these factors and carbon emissions based on the Extra-trees model. Results show that energy consumption (ENC) remains the root cause of increased carbon emissions among domestic economic factors, although government intervention (GOV) and digital finance (DIG) can significantly reduce it. Next, among the external economic and policy uncertainty factors, foreign direct investment (FDI) and economic policy uncertainty (EPU) are important factors influencing carbon emissions, and the partial dependence plots (PDPs) confirm the pollution haven hypothesis and also reveal the role of EPU in reducing carbon emissions. The heterogeneity of factors affecting carbon emissions is also analyzed under different city sizes, and it is found that ENC is a common driving factor in cities of different sizes, but there are some differences. Finally, appropriate policy recommendations are proposed by us to help China move rapidly towards a green and sustainable development path.

Robust prediction of colorectal cancer via gut microbiome 16S rRNA sequencing data.

Introduction. The study addresses the challenge of utilizing human gut microbiome data for the early detection of colorectal cancer (CRC). The research emphasizes the potential of using machine learning techniques to analyze complex microbiome datasets, providing a non-invasive approach to identifying CRC-related microbial markers.Hypothesis/Gap Statement. The primary hypothesis is that a robust machine learning-based analysis of 16S rRNA microbiome data can identify specific microbial features that serve as effective biomarkers for CRC detection, overcoming the limitations of classical statistical models in high-dimensional settings.Aim. The primary objective of this study is to explore and validate the potential of the human microbiome, specifically in the colon, as a valuable source of biomarkers for colorectal cancer (CRC) detection and progression. The focus is on developing a classifier that effectively predicts the presence of CRC and normal samples based on the analysis of three previously published faecal 16S rRNA sequencing datasets.Methodology. To achieve the aim, various machine learning techniques are employed, including random forest (RF), recursive feature elimination (RFE) and a robust correlation-based technique known as the fuzzy forest (FF). The study utilizes these methods to analyse the three datasets, comparing their performance in predicting CRC and normal samples. The emphasis is on identifying the most relevant microbial features (taxa) associated with CRC development via partial dependence plots, i.e. a machine learning tool focused on explainability, visualizing how a feature influences the predicted outcome.Results. The analysis of the three faecal 16S rRNA sequencing datasets reveals the consistent and superior predictive performance of the FF compared to the RF and RFE. Notably, FF proves effective in addressing the correlation problem when assessing the importance of microbial taxa in explaining the development of CRC. The results highlight the potential of the human microbiome as a non-invasive means to detect CRC and underscore the significance of employing FF for improved predictive accuracy.Conclusion. In conclusion, this study underscores the limitations of classical statistical techniques in handling high-dimensional information such as human microbiome data. The research demonstrates the potential of the human microbiome, specifically in the colon, as a valuable source of biomarkers for CRC detection. Applying machine learning techniques, particularly the FF, is a promising approach for building a classifier to predict CRC and normal samples. The findings advocate for integrating FF to overcome the challenges associated with correlation when identifying crucial microbial features linked to CRC development.

Integrating hydrological parameters in wildfire risk assessment: a machine learning approach for mapping wildfire probability

Abstract The increasing occurrence of catastrophic wildfire across the globe threatens public health, community safety, ecosystem functioning, and biodiversity resilience. Wildfire risk is closely connected to shifting climatic trends and their impacts on fuel availability and flammability. Although previous research has explored the connection between meteorological conditions and wildfire probabilities, there remains a substantial gap in understanding the influence of hydrologic drivers, such as groundwater recharge, on wildfire dynamics. Both short- and long-term variations in these variables are crucial in shaping fuel conditions, and significant changes can create environments more prone to severe wildfires. This study focuses on Santa Barbara County to examine the connection between wildfire probability and various environmental factors, including meteorological and hydrological data from 1994 to 2021, topography, vegetation, and proximity to road. Using a random forest (RF) machine learning model and fine-scale data (270 m resolution) we achieved high predictive accuracy in identifying wildfire probability. Our findings confirm the important roles of short-term meteorological conditions, such as mean precipitation 12 months and relative humidity 1 month before a wildfire event, in predicting wildfire occurrence. In addition, our results emphasize the critical contribution of long-term hydrological components, such as mean deviation from the historical normal in actual evapotranspiration and recharge in the years preceding the fire, in influencing wildfire probability. Partial dependence plots from our RF model revealed that both positive and negative deviations of these hydrological variables can increase the likelihood of wildfire by controlling fuel water availability and productivity. These findings are particularly relevant given the increasing extreme weather patterns in southern California, significantly affecting water availability and fuel conditions. This study provides valuable insights into the complex interactions between wildfire occurrence and hydrometeorological conditions. Additionally, the resulting wildfire probability map, can aid in identifying high-risk areas, contributing to enhanced mitigation planning and prevention strategies.

DNN-metamodeling and fragility estimate of high-rise buildings with outrigger systems subject to seismic loads

This paper proposed deep neural networks (DNNs) for the dynamic response of high-rise buildings with one-outrigger systems under two types of seismic hazards. Using an existing database, the hyperparameters for the architecture of the DNNs are determined finding a trade-off between accuracy and complexity. The performance of the proposed DNNs is compared with two metamodels in the literature, a probabilistic demand model and a kriging metamodel. Partial dependence plot and SHapley Additive exPlanations are used for the explanation of the marginal effect of each feature on the predicted outcome of a neural network from the perspective of global and local agnostic. Considering the uncertainty in the input features, the DNNs are then used to formulate fragility estimates for example high-rise buildings with three types of outrigger systems. The model-agnostic analysis suggests that the DNN targeting the inter-story drift and top acceleration shows extremely high sensitivity to variation in the seismic hazard features, earthquake magnitude, and rupture distance. The fragility curves objectively quantify the reliability of buildings with each of the three outrigger systems and show the effectiveness of damped outrigger systems in reducing fragilities.

Machine Learning Model Based on Prognostic Nutritional Index for Predicting Long-Term Outcomes in Patients With HCC Undergoing Ablation.

To develop multiple machine learning (ML) models based on the prognostic nutritional index (PNI) and determine the optimal model for predicting long-term survival outcomes in hepatocellular carcinoma (HCC) patients after local ablation. From January 2009 to December 2019, we analyzed data from 848 primary HCC patients who underwent local ablation. ML models were constructed and evaluated using the concordance index (C-index), concordance-discordance area under curve (C/D AUC), and Brier scores. The optimal ML model was interpreted using the partial dependence plot (PDP) and SHapley Additive exPlanations (SHAP) framework. Additionally, the prognostic performance of our model was compared with other models. Alkaline phosphatase, preoperation alpha-fetoprotein level, PNI, tumor number, and tumor size were identified as independent prognostic factors for ML model construction. Among the 19 ML algorithms tested, the Aorsf model showed superior performance in both the training cohort (C/D AUC: 0.733; C-index: 0.736; Brier score: 0.133) and validation cohort (C/D AUC: 0.713; C-index: 0.793; Brier score: 0.117). The time-dependent AUC of the Aorsf model for predicting overall survival was as follows: 1-, 3-, 5-, 7-, and 9-year were 0.828, 0.765, 0.781, 0.817, and 0.812 in the training cohort, 0.846, 0.859, 0.824, 0.845, and 0.874 in the validation cohort, respectively. The PDP and SHAP algorithms were employed for visual interpretation. Furthermore, time-AUC and decision curve analysis demonstrated that the Aorsf model provided superior clinical benefits compared to other models. The PNI-based Aorsf model effectively predicts long-term survival outcomes after ablation therapy, making a significant contribution to HCC research by improving surveillance, prevention, and treatment strategies.

Hydrothermal bio-oil yield and higher heating value of high moisture and lipid biomass: Machine Learning modeling and feature response behavior analysis

The yield and higher heating value (HHV) of bio-oil products are significant performance parameters for the hydrothermal conversion of high-water and high-lipid biomass. Machine learning (ML) modeling prediction is a fast and convenient means of obtaining performance parameters. An informative dataset with 243 samples was prepared, and two highly adapted ML algorithms were used: Random Forest (RF) and Extreme Gradient Boosting Tree (XGBoost). It is interesting to note that the developed ML models demonstrated great prediction ability; for example, the regression coefficient () of the XGBoost model for yield and HHV prediction was as high as 0.942 and 0.940, respectively. Furthermore, partial dependence plots and SHAP (SHapley Additive exPlanations) interpretability methodologies were adopted to address the main contributions of the feature identification and response behavior analysis of the features. The results demonstrated that the biomass composition had the greatest effect on bio-oil yield, with fat contributing up to 40%. In contrast, the elemental composition had the most significant effect on the HHV of bio-oil. Notably, hydrogen content affected the HHV of up to 4.5 units. The interaction response behavior showed that the interaction of the process parameters with feedstock properties was most common and significant. The information obtained from the response mechanism can be used to enhance the subsequent hydrothermal fuel preparation process for bio-oils.

Innovative hybrid machine learning models for estimating the compressive strength of copper mine tailings concrete

The growing demand for copper and related materials in various industries is driving increased copper mining globally. This surge presents a substantial challenge in managing and responsibly disposing of large volumes of copper mine tailings (CMT). Incorporating CMT as supplementary cementitious materials (SCMs) in concrete addresses two significant environmental challenges simultaneously: reducing the accumulation of CMT waste in landfills and lowering the carbon footprint by reducing cement usage. The investigation into recycling CMT as a cement substitute involves a thorough assessment of its impact on the compressive strength (CS) of concrete. This research introduces innovative hybrid machine learning (ML) models for estimating the CS of CMT concrete, aiming to streamline strength assessment processes and save valuable resources. The method involves integrating features from large public datasets with the limited available data on the CS of CMT concrete. Support vector regression (SVR) was combined with advanced optimization techniques: firefly algorithm (FFA), grey wolf optimization (GWO) and particle swarm optimization (PSO) to create new hybrid models for forecasting the CS of CMT concrete. Additionally, traditional ML techniques like decision tree (DT) and random forest (RF) were used to compare with these SVR-based hybrids. All three hybrid models demonstrated strong performance, with SVR-FFA emerging as the most effective among them. Notably, SVR-FFA achieved the greatest R² score of 0.96, indicating superior predictive accuracy compared to SVR-PSO (0.92) and SVR-GWO (0.90). Additionally, the DT model attained an R² score of 0.88, while the RF model achieved an R² score of 0.84. Moreover, the SHapley additive exPlanations (SHAP) and partial dependence plots (PDP) analyses underscore the positive effects of curing age, cement, blast furnace slag, and superplasticizer on the CS of CMT concrete. A graphical user interface was developed for predicting the CS of CMT concrete, allowing for instant predictions without the need for conducting experiments.

Machine learning assisted prediction of specific surface area and nitrogen content of biochar based on biomass type and pyrolysis conditions

Predicting and optimizing the physicochemical properties of biochar is crucial for its applications. The characteristics of biomass and pyrolysis conditions are the main factors influencing these properties. However, the numerous components of biomass and the pyrolysis conditions contribute to the substantial challenge in predicting the physicochemical properties, particularly the specific surface area and the nitrogen content of biochar. In this work, machine learning methods including random forest (RF), gradient boosting decision tree (GBDT) and extreme gradient boosting (XGB) (all with R2 exceeding 0.97) were used to predict and analyze specific surface area of biochar (SSA), N content of biochar (N-char), and yield of biochar (Yield-char). Compositions of biomass and pyrolysis conditions were selected as input variables. The partial dependence plot analysis showed the impact way of each influential factor on the target variable and the interactions among these factors in the pyrolysis process. The feature importance of these models indicated that the influencing factors toward predicting three targets (sorted by importance) were specified as follows: pyrolysis temperature, nitrogen content, and fixed carbon for Yield-char; N and ash for N-char; ash and pyrolysis temperature for SSA. This work provided new insights for understanding pyrolysis process of biomass.

The application of machine learning for identifying frailty in older patients during hospital admission

BackgroundEarly identification of frail patients and early interventional treatment can minimize the frailty-related medical burden. This study investigated the use of machine learning (ML) to detect frailty in hospitalized older adults with acute illnesses.MethodsWe enrolled inpatients of the geriatric medicine ward at Taichung veterans general hospital between 2012 and 2022. We compared four ML models including logistic regression, random forest (RF), extreme gradient boosting, and support vector machine (SVM) for the prediction of frailty. The feature window as well as the prediction window was set as half a year before admission. Furthermore, Shapley additive explanation plots and partial dependence plots were used to identify Fried’s frailty phenotype for interpreting the model across various levels including domain, feature, and individual aspects.ResultsWe enrolled 3367 patients. Of these, 2843 were frail. We used 21 features to train the prediction model. Of the 4 tested algorithms, SVM yielded the highest AUROC, precision and F1-score (78.05%, 94.53% and 82.10%). Of the 21 features, age, gender, multimorbidity frailty index, triage, hemoglobin, neutrophil ratio, estimated glomerular filtration rate, blood urea nitrogen, and potassium were identified as more impactful due to their absolute values.ConclusionsOur results demonstrated that some easily accessed parameters from the hospital clinical data system can be used to predict frailty in older hospitalized patients using supervised ML methods.

Quantitative assessment of interplay between urbanization dynamics and land surface temperature variations using generalized additive model coupled PDP for sustainable urban planning and management.

The mountainous region of Asir is experiencing rapid and unsystematic urbanization leading to an increase in land surface temperatures (LST), which poses a challenge to human well-being and ecological balance. Therefore, it is necessary to study the interaction between land use and land cover (LULC)-induced urbanization and LST using advanced geostatistical techniques. In addition, understanding the urbanization process and urban density is essential for effective urban planning and management. The aim of this study was to investigate the interaction between the urbanization process, urban density and the associated LST. Using the Random Forest Algorithm, LULC mapping was conducted for the years 1990, 2000 and 2020. Metrics such as land cover change rate (LCCR), land cover index (LCI), landscape expansion index (LEI), mean landscape expansion index (MLEI) and area-weighted landscape expansion index (AWLEI) were used to understand urbanization processes and LULC changes. Convolutional kernels were used to model urban density, and the mono-window algorithm was applied to analyse LST in the selected years. In addition, the study assessed the Surface Urban Heat Island (SUHI) contribution index to LULC and used Generalized Additive Models (GAMs) in conjunction with Partial Dependence Plots (PDPs) to understand the relationship between urbanization processes, urban density and LST. In a detailed 30-year study, the application of the RF algorithm showed significant shifts in LULC with an overall validation accuracy of over 85%. Urban areas grew dramatically from 69.40 km2 in 1990 to 338.74 km2 in 2020, while water areas decreased from 1.51 to 0.54 km2. Dense vegetation increased from 43.36 to 52.22 km2, indicating positive ecological trends. The LST analysis showed a general warming, with the mean LST increasing from 40.51°C in 1990 to 46.73°C in 2020 and the highest temperature category (50-60°C) increasing from 0.78 to 33.35 km2. The built-up area of cities tripled between 1990 and 2020, with the Landscape Expansion Index reflecting significant growth in suburban areas. The modeling of urban density shows increasing urbanization in the centre, which will expand significantly to the east by 2020. The contribution of LULC to LST and the Urban Heat Island (SUHI) effect was evident, with built-up areas showing a constant temperature increase. GAMs confirmed a statistically significant relationship between urban density and LST, with different effects for different types of urban expansion. This comprehensive study quantitatively sheds light on the complicated dynamics of urbanization, land cover change and temperature variation and provides important insights for sustainable urban development.

Integration of interpretable machine learning and environmental magnetism elucidates reduction mechanism of bioavailable potentially toxic elements in lakes after monsoon

Little information is available on the influence of substantial precipitation and particulate matter entering during the monsoon process on the release of potentially toxic elements (PTEs) into lake sediments. Sediments from a typical subtropical lake across three periods, pre-monsoon, monsoon, and post-monsoon, were collected to determine the chemical forms of 12 PTEs (As, Cd, Co, Cr, Cu, Fe, Hg, Pb, Mn, Ni, Sb, and Zn), magnetic properties, and physicochemical indicators. Feature importance, Shapley additive explanations, and partial dependence plots were used to explore the factors influencing bioavailable PTEs. The proportion of bioavailable forms of PTEs decreased from 3.85 % (Cd) to 87.84 % (Hg) after the monsoon. Gradient extreme boosting demonstrated robust fitting accuracy for the prediction of the bioavailable forms of the 12 PTEs (R2 > 0.84). Shapley additive explanations identified that the bioavailable forms were influenced by the total PTE concentrations, wind, shortwave radiation, and particle inputs (25.1 %–88.5 % for total importance), either individually or in combination. The partial dependence plots highlighted the influence thresholds of background values and anthropogenic factors on the bioavailable forms of PTEs. Changes in environmental properties could indicate the process of external sediment influx into lakes. The optimized model combined with magnetic parameters showed strong performance in other cases (coefficient of determination>0.58), confirming the ubiquitous decrease in bioavailable forms of PTEs in sediments across subtropical lakes after monsoons.

Deciphering the cytotoxicity of micro- and nanoplastics in Caco-2 cells through meta-analysis and machine learning

Plastic pollution, driven by micro- and nanoplastics (MNPs), poses a major environmental threat, exposing humans through various routes. Despite human colorectal adenocarcinoma Caco-2 cells being used as an in vitro model for studying the intestinal epithelium, uncertainties linger about MNPs harming these cells and the factors influencing adverse effects. Addressing this lacuna, our study aimed to elucidate the pivotal MNP parameters influencing cytotoxicity in Caco-2 cells, employing meta-analysis and machine learning techniques for quantitative assessment. Initial scrutiny of 95 publications yielded 17 that met the inclusion criteria, generating a dataset of 320 data points. This dataset underwent meticulous stratification based on polymer type, exposure time, polymer size, MNP concentration, and biological assays utilised. Subsequent dose-response curve analysis revealed moderate correlations for selected subgroups, such as the (3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide) MTT biological assay and exposure time exceeding 24 h, with coefficient of determination (R2) values of 0.50 (p-value: 0.0065) and 0.60 (p-value: 0.0018) respectively. For the aforementioned two subgroups, the MNP concentrations surpassing 10 μg/mL led to diminished viability of Caco-2 cells. Notably, we observed challenges in employing meta-analysis to navigate this multidimensional MNP dataset. Leveraging a random forest model, we achieved improved predictive performance, with R2 values of 0.79 and a root mean square error (RMSE) of 0.14 for the prediction of the Log Response Ratio on the test set. Model interpretation indicated that size and concentration are the principal drivers influencing Caco-2 cell cytotoxicity. Additionally, the partial dependence plot illustrating the relationship between the size of MNPs and predicted cytotoxicity reveals a complex pattern. Our study provides crucial insights into the health impacts of plastic pollution, informing policymakers for targeted interventions, thus contributing to a comprehensive understanding of its human health consequences.