Random Forest Machine Learning Research Articles

Risk factors of pancreatic fistula after distal pancreas resections: an observational retrospective study

BACKGROUND: The main complication of surgical pancreas interventions is the pancreatic fistula (PF) — abnormal secretion of pancreatic secretions into the abdominal cavity, which can lead to more severe complications. At the same time, some predictors of PF remain poorly understood. AIM: To investigate the risk factors for the development of PF after distal pancreatic resection. MATERIALS AND METHODS: The study included 40 patients who underwent distal resection. The influence of various clinical parameters on the development of PF after surgery was evaluated. Methods of mathematical modeling and correlation analysis were used. Mathematical modeling was carried out using the Random forest machine learning algorithm, and the indicator of the relative importance of the “importance” parameters was evaluated. RESULTS: According to the Random forest model, on day 1 after surgery, the most significant predictors of the development of PF were: the volume of neoplasm, age and stage of the oncological process, for which “importance” was 53.2, 13.7 and 12.5 (AUC ROC=62%); on day 3–5, “importance” was 61.7, 11.5 and 5.2 (AUC ROC=79%). An increase in the concentration of pancreatic amylase in blood plasma for 2–3 and 3–5 days correlated with its increase in the drainage discharge for 5-7 days after the intervention (r=0.48 and 0.76; p 0.05). A correlation was found between the level of amylase in the drainage discharge on 3–5 days after the intervention and the level of leukocytes according to the general blood test (r=0.62, p 0.05). CONCLUSION: An increase in plasma amylase levels is the main risk factor for the development of PF. An increase in the concentration of amylase in the drainage discharge can be considered as a potential risk factor for the development of pancreatic fistulas with clinical manifestations in the following days. The stage of the oncological process, the size of the neoplasm and the age of the patient are also risk factors.

Smart hotspot detection using geospatial artificial intelligence: A machine learning approach to reduce flood risk

This study employs Geospatial Artificial Intelligence (GeoAI) and the Random Forest Machine Learning (ML) algorithm to enhance flood hazard assessments in Portugal. It utilizes NASA's LP DAAC (2023) Digital Elevation Model (DEM) and slope data from EPIC WEBGIS PORTUGAL DATA, offering detailed topographical insights for environmental planning. Additionally, it incorporates data on proximity to water bodies from the Portuguese Environment Agency and the European Environment Agency, and soil characteristics from EPIC WEBGIS PORTUGAL DATA, facilitating a thorough examination of flood risks. This approach prioritizes long-term land features over short-term weather patterns, providing a comprehensive understanding of flood vulnerability. The study processes data at a 1 km x 1 km resolution, adapting TIFF maps for compatibility with the Random Forest model. The produced flood hazard maps identify potential flood hotspots at both national and city levels, crucial for urban planning. These maps aid in assessing the vulnerability of key infrastructure and assets, such as transport networks and buildings. The research highlights the importance of integrating additional data on assets and socioeconomic factors to enhance urban resilience. It sets the stage for future research aimed at improving predictive accuracy and underscores the necessity of extensive geospatial analytics in managing infrastructure risks.

Modeling forest canopy structure and developing a stand health index using satellite remote sensing

Biotic and abiotic disturbances modify tree structure and degrade stand health. Accurate geospatial data on stand structure is important for monitoring tree growth, forest health, progression and severity of diseases and pests, and estimating resilience to climate stress. The live crown ratio (LCR) of trees serves as a key health indicator but has been understudied at the landscape level using remote sensing data. This study generated the leaf area index (LAI) and a novel spatial layer of LCR at site and landscape scales using a combination of satellite data and ground observations. We conducted field surveys to collect plot-level (10 m × 10 m) data in four eastern white pine (EWP; Pinus strobus L.)-dominated sites in the state of Maine, USA. The plot-level data were used to develop regression models for LAI and LCR estimation using microwave (Sentinel-1) and optical (Sentinel-2) remote sensing data and applying the Random Forest (RF) and Support Vector Machine (SVM) machine learning algorithms. The RF model showed higher prediction accuracy than the SVM model at the site level. Moreover, the prediction accuracy at the site and landscape levels were comparable for LAI (R2 > 0.76) and LCR (R2 > 0.71) using the RF model. Furthermore, the predicted LAI and LCR were integrated with canopy height and stand density to develop a novel health index map for EWP. The resulting health index map successfully delineated patches representing various health categories. Forestry practitioners and decision-makers can use the derived health index map and intermediate spatial data layers (LAI and LCR) to guide stand management. The developed framework can potentially be applied to other coniferous and broadleaved species for remote sensing-based LCR estimation and forest health assessment upon further studies and verification.

Landscape and climatic factors shaping mosquito abundance and species composition in southern Spain: A machine learning approach to the study of vector ecology

Vector-borne diseases pose significant challenges to public health, with mosquitoes acting as crucial vectors for pathogens globally. This study explores the interaction between environmental and climate factors, investigating their influence on the abundance and species composition of mosquitoes in southwestern Spain, a region endemic to several mosquito-borne diseases.Using comprehensive field data from 2020, we analysed mosquito abundance and species richness alongside remote sensing variables and modeling techniques, including the machine learning Random Forest. We collected 5859 female mosquitoes representing 13 species. Non-linear correlations were observed between mosquito abundance and climatic variables, notably temperature and rainfall. Extremely high temperatures correlated with a decrease in mosquito abundance, while accumulated rainfall in the three weeks preceding sampling positively impacted mosquito abundance by providing breeding habitats. A positive correlation between Normalized Difference Vegetation Index (NDVI) and mosquito metrics was also found, aligning with prior studies highlighting vegetation's role shaping mosquito habitats. Interestingly, a negative relationship was observed between mosquito species richness and autumn NDVI. Additionally, wind speed negatively affected mosquito species richness.This research provides valuable insights into the ecological determinants of mosquito abundance and species composition in a Mediterranean climate. These findings are crucial for understanding disease transmission dynamics and improving vector control strategies. By integrating climatic characteristics into public health interventions, management measures can become more targeted and efficient, especially during periods of heightened temperature.

Slick: Modeling a Universe of Molecular Line Luminosities in Hydrodynamical Simulations

We present slick (the Scalable Line Intensity Computation Kit), a software package that calculates realistic CO, [C i], and [C ii] luminosities for clouds and galaxies formed in hydrodynamic simulations. Built on the radiative transfer code despotic, slick computes the thermal, radiative, and statistical equilibrium in concentric zones of model clouds, based on their physical properties and individual environments. We validate our results by applying slick to the high-resolution run of the Simba simulations, testing the derived luminosities against empirical and theoretical/analytic relations. To simulate the line emission from a universe of emitting clouds, we have incorporated random forest machine learning (ML) methods into our approach, allowing us to predict cosmologically evolving properties of CO, [C i], and [C ii] emission from galaxies such as luminosity functions. We tested this model in 100,000 gas particles, and 2500 galaxies, reaching an average accuracy of ∼99.8% for all lines. Finally, we present the first model light cones created with realistic and ML-predicted CO, [C i], and [C ii] luminosities in cosmological hydrodynamical simulations, from z = 0 to z = 10.

A Stackelberg game framework for promoting battery-powered construction machinery through government subsidies

Diesel-powered construction machinery is a major source of harmful pollutants and carbon emissions. Technological advancements have made battery-powered construction machines (BPCMs) increasingly viable in operations. However, the high purchase prices of BPCMs represent a huge obstacle for contractors to embrace these low-emission alternatives. To facilitate the sustainable development of the construction industry, it is imperative for the government to implement effective subsidy allocation policies to promote the electrification of construction machinery. This paper proposes a Stackelberg game framework for optimal subsidy allocation. In the proposed framework, the government decides the subsidy amount offered to each type of BPCMs to minimize both pollutant emissions and carbon emissions. The contractors can then observe the government’s decision and make their optimal decisions regarding the purchase, operation, and replacement of construction machines accordingly to minimize their total costs. Contractors’ decisions in turn influence the government’s decision. Such an intricate framework has many appealing properties, which are analyzed in depth to provide useful managerial insights. Additionally, the design of effective subsidy policies by the government depends on a precise prediction of the contractors’ demand for construction machinery. To this end, a random forest machine learning model is developed. Real data were collected for model construction and testing. Statistical results and industrial comments show the high quality of our predictions.

Primary Hyperhidrosis and Sensitive Skin: Exploring the Link with Predictive Machine Learning-Based Classification Models.

Primary hyperhidrosis (PHH) is a disorder of excessive sweating caused by aberrant cholinergic signaling. Sensitive skin (SS) is a condition of subjective cutaneous hyperreactivity to innocuous stimuli, impacting 40% to 70% of the population. SS is exacerbated by sweat, stress, and heat, suggesting that cholinergic stimulation may contribute to SS flares. To survey PHH sufferers to assess hyperhidrosis (HH) and SS symptom burden. An International Review Board (IRB)-exempt survey was disseminated by the International Hyperhidrosis Society. A predictive classification model for SS was built using random forest machine learning algorithms. Of the 637 respondents with PHH, 89% reported SS; and there was a significant association between HH and SS severity scores. Importantly, SS occurred on body sites affected and unaffected by HH. Predictive modeling designated Sensitive Scale-10 (SS-10), a validated questionnaire to gauge SS severity, to be the most helpful in predicting SS in this cohort. Self-reported data. These data are the first to propose and support a relationship between SS and HH. SS occurred with greatest frequency at HH-afflicted body sites, but also occurred on unaffected sites, suggesting that sweat is not the sole causative link. Future work can explore cholinergic signaling as a potential link between these conditions. Screening HH patients for SS may be warranted. J Drugs Dermatol. 2024;23(10):882-888. doi:10.36849/JDD.8461.

Predictive Cultivation: Integrating Meteorological Data and Machine Learning for Enhanced Crop Yield Forecast

Agriculture is a key component of Telangana’s economy, and greater performance in this sector is crucial for inclusive growth. A central challenge is yielding estimation to predict crop yields before harvesting. This paper addresses this challenge with machine learning approaches includes Naive Bayes, KNN and Random Forest. The parameters considered for model testing are crop, season, rainfall and location. This paper includes a case study of Telangana with the help of Telangana weather data set to provide analysis on the key factors like overall rainfall recorded with respect to each Mandal, overall seasonal yield in selected years, seasonal yield of major crops like Bengal gram, groundnut and maize, and overall yield in two different agricultural seasons: rabi and kharif. Random forest machine learning model produces highest accuracy of 99.32% when compared with other process models.

Return management in e-commerce firms: A machine learning approach to predict product returns and examine variables influencing returns

The advent of technology with internet penetration has led to exponential growth in sales through online mediums worldwide. Although the growth has increased sales and created opportunities for many firms and sellers to flourish, it also brings in the challenge of customer returns, making it a formidable challenge for businesses. For effective return handling and control, predicting product returns correctly is of utmost importance for firms. The study comprehensively examines the return management process, elucidating its multiple features and characteristics, focusing on the types of returns and the involvement of various stakeholders. Logistic regression, random forest, and gradient-boosting machine learning techniques were used on a data set to scrutinize critical variables influencing returns. Comparing the models using the ROC curve and AUC, gradient boosting performed better in predicting returns than other models. The study identifies significant variables affecting e-commerce return rates, such as total order amount, product category, and cash on delivery charge, an additional fee charged on this mode of payment. Managerial insights recommend that firms conduct due diligence and thorough assessments before introducing products from specific vendors. Quality improvement in certain categories of products, incentivizing card payments, and nominalizing the cash on delivery fees, which could be seen as burdensome by customers, are some of the steps firms need to reduce returns. Strategies like confirming weekend orders and updating replenishment times regularly by vendors can help firms mitigate returns effectively. The study contributes to the academic understanding of return management in e-commerce and offers actionable insights for practitioners.

European Union crop map 2022: Earth observation’s 10-meter dive into Europe’s crop tapestry

To provide the information needed for a detailed monitoring of crop types across the European Union (EU), we present an advanced 10-metre resolution map for the EU and Ukraine with 19 crop types for 2022, updating the 2018 version. Using Earth Observation (EO) and in-situ data from Eurostat’s Land Use and Coverage Area Frame Survey (LUCAS) 2022, the methodology included 134,684 LUCAS Copernicus polygons, Sentinel-1 and Sentinel-2 satellite imagery, land surface temperature and a digital elevation model. Based on this data, two classification layers were developed using a Random Forest machine learning approach: a primary map and a gap-filling map to address cloud-covered gaps. The combined maps, covering 27 EU countries, show an overall accuracy of 79.3% for seven major land cover classes and 70.6% for all 19 crop types. The trained model was used to derive the 2022 map for Ukraine, demonstrating its robustness even in regions without labelled samples for model training.

Between Two Worlds: Investigating the Intersection of Human Expertise and Machine Learning in the Case of Coronary Artery Disease Diagnosis.

Coronary artery disease (CAD) presents a significant global health burden, with early and accurate diagnostics crucial for effective management and treatment strategies. This study evaluates the efficacy of human evaluators compared to a Random Forest (RF) machine learning model in predicting CAD risk. It investigates the impact of incorporating human clinical judgments into the RF model's predictive capabilities. We recruited 606 patients from the Department of Nuclear Medicine at the University Hospital of Patras, Greece, from 16 February 2018 to 28 February 2022. Clinical data inputs included age, sex, comprehensive cardiovascular history (including prior myocardial infarction and revascularisation), CAD predisposing factors (such as hypertension, dyslipidemia, smoking, diabetes, and peripheral arteriopathy), baseline ECG abnormalities, and symptomatic descriptions ranging from asymptomatic states to angina-like symptoms and dyspnea on exertion. The diagnostic accuracies of human evaluators and the RF model (when trained with datasets inclusive of human judges' assessments) were comparable at 79% and 80.17%, respectively. However, the performance of the RF model notably declined to 73.76% when human clinical judgments were excluded from its training dataset. These results highlight a potential synergistic relationship between human expertise and advanced algorithmic predictions, suggesting a hybrid approach as a promising direction for enhancing CAD diagnostics.

Enhanced Monitoring of Sub-Seasonal Land Use Dynamics in Vietnam’s Mekong Delta through Quantile Mapping and Harmonic Regression

In response to economic and environmental challenges like sea-level rise, salinity intrusion, groundwater extraction, sand mining, and sinking delta phenomena, the demand for solutions to adapt to changing conditions in riverine environments has increased significantly. High-quality analyses of land use and land cover (LULC) dynamics play a critical role in addressing these challenges. This study introduces a novel high-spatial resolution satellite-based approach to identify sub-seasonal LULC dynamics in the Mekong River Delta (MRD), employing a three-year (2021–2023) Sentinel-1 and Sentinel-2 satellite data time series. The primary obstacle is discerning detailed vegetation dynamics, particularly the seasonality of rice crops, answered through quantile mapping, harmonic regression with Fourier transform, and phenological metrics as inputs to a random forest machine learning classifier. Due to the substantial data volume, Google’s cloud computing platform Earth Engine was utilized for the analysis. Furthermore, the study evaluated the relative significance of various input features. The overall accuracy of the classification is 82.6% with a kappa statistic of 0.81, determined using comprehensive reference data collected in Vietnam. While the purely pixel-based approach has limitations, it proves to be a viable method for high-spatial resolution satellite image time series classification of the MRD.

Estimating Breakfast Characteristics Using Continuous Glucose Monitoring and Machine Learning in Adults With or at Risk of Type 2 Diabetes

Background: Continuous glucose monitoring (CGM) systems allow detailed assessment of postprandial glucose responses (PPGR), offering new insights into food choices’ impact on dysglycemia. However, current approaches to analyze PPGR using a CGM require manual meal logging, limiting the scalability of CGM-driven applications like personalized nutrition and at-home diabetes risk assessment. Objective: We propose a machine learning (ML) framework to automatically identify and characterize breakfast-related PPGRs from CGM profiles in adults at risk of or living with noninsulin-treated type 2 diabetes (T2D). Methods: Our PPGR estimation framework uses a random forest ML algorithm trained on 15 adults without diabetes who wore a CGM for up to four weeks. The algorithm performance was evaluated on a held-out subset of the participants’ CGM data as well as on an external validation data set of 36 individuals at risk for or with noninsulin-treated T2D. Results: Our algorithm’s estimations of breakfast PPGRs displayed no statistically significant differences to annotated PPGRs, in terms of incremental area under the curve and glucose rise ( P > .05 for both data sets), while a small difference in prebreakfast glucose was found in the nondiabetes data set ( P = .005) but not in the validation T2D data set ( P = .18). Conclusions: We designed an ML framework to automatically estimate the timing of meal events from CGM data in individuals without diabetes and in individuals at risk or with T2D. This could provide a more scalable approach for analyzing postprandial glycemia, increasing the feasibility of CGM-based precision nutrition and diabetes risk assessment applications.

Predicting Primary Graft Dysfunction in Lung Transplantation: Machine Learning-Guided Biomarker Discovery.

There is an urgent need to better understand the pathophysiology of primary graft dysfunction (PGD) so that point-of-care methods can be developed to predict those at risk. Here we utilize a multiplex multivariable approach to define cytokine, chemokines, and growth factors in patient-matched biospecimens from multiple biological sites to identify factors predictive of PGD. Biospecimens were collected from patients undergoing bilateral LTx from three distinct sites: donor lung perfusate, post-transplant bronchoalveolar lavage (BAL) fluid (2h), and plasma (2h and 24h). A 71-multiplex panel was performed on each biospecimen. Cross-validated logistic regression (LR) and random forest (RF) machine learning models were used to determine whether analytes in each site or from combination of sites, with or without clinical data, could discriminate between PGD grade 0 ( n = 9) and 3 ( n = 8). Using optimal AUROC, BAL fluid at 2h was the most predictive of PGD (LR, 0.825; RF, 0.919), followed by multi-timepoint plasma (LR, 0.841; RF, 0.653), then perfusate (LR, 0.565; RF, 0.448). Combined clinical, BAL, and plasma data yielded strongest performance (LR, 1.000; RF, 1.000). Using a LASSO of the predictors obtained using LR, we selected IL-1RA, BCA-1, and Fractalkine, as most predictive of severe PGD. BAL samples collected 2h post-transplant were the strongest predictors of severe PGD. Our machine learning approach not only identified novel cytokines not previously associated with PGD, but identified analytes that could be used as a point-of-care cytokine panel aimed at identifying those at risk for developing severe PGD.

Analysis and performance prediction of a building integrated photovoltaic thermal system with and without phase change material

Increasing energy demand in buildings in the last decade has led to more efficient usage of renewable energy. Building integrated photovoltaic/thermal systems is one of the effective ways of reducing energy consumption in buildings by providing heating, cooling, ventilation, hot water, and air. The enhancement of photovoltaic cell temperature reduces the electrical efficiency of photovoltaic panel systems. In this regard, the effects of cooling with air and phase change material layer are investigated in this study. The simulation of BIPV/T, PV, and heat pump was done in TRNSYS. As the main novelty of recent simulation, for the calculations of phase change material layer, MATLAB coupled with the TRNSYS, and Python software was used for the prediction process under the weather conditions of Sharurah, Saudi Arabia for 2022. Also, the performance of a building integrated photovoltaic/thermal system with a phase change material layer is predicted for 2024 and 2025 by the Random Forest machine learning method. The R squared, Root Mean Squared Error, and Mean Absolute Error are utilized for checking the accuracy of the model. Also, for more certainty, a validation was done on the simulated results of TRNSYS and the results showed high conformity of data. The simulated data for the first six months of 2022 showed that the maximum cooling effect happened in June with 25.47 °C and a 37.87 % reduction in photovoltaic cell temperature. Also, the best electrical efficiency improvement was 1.26 % while the thermal efficiency ranged from 29.65 % to 59.40 %, respectively. The predicted data was confirmed by R2, RMSE, and MAE equal to 96.64 %, 0.06 and 0.04. According to the simulation and predictions done, the annual electrical production of the building is achieved. The results indicate that, for the simulated system, about 29.7 %, 29.8 %, and 29.6 % of total energy utilization could be supplied by BIPV/T cooled with phase change material in 2022, 2024, and 2025, respectively.

Differential Classification of Dengue, Zika, and Chikungunya Using Machine Learning—Random Forest and Decision Tree Techniques

Differential Classification of Dengue, Zika, and Chikungunya Using Machine Learning—Random Forest and Decision Tree Techniques

The Impact of Starting Positions and Breathing Rhythms on Cardiopulmonary Stress and Post-Exercise Oxygen Consumption after High-Intensity Metabolic Training: A Randomized Crossover Prospective Study.

Background: The exploration of optimizing cardiopulmonary function and athletic performance through high-intensity metabolic exercises (HIMEs) is paramount in sports science. Despite the acknowledged efficacy of HIMEs in enhancing cardiopulmonary endurance, the high metabolic stress imposed on the cardiopulmonary system, especially for amateurs, necessitates a scaled approach to training. Objective: The aim of this study is to ascertain whether adjustments in the initiation posture and the adoption of an appropriate breathing strategy can effectively mitigate the cardiopulmonary stress induced by HIMEs without compromising training efficacy. Methods: Twenty-two subjects were recruited into this study. The post-exercise heart rate (PHR) and post-exercise oxygen consumption rate (POCR) were collected within 30 min after exercise. A two-way ANOVA, multi-variable Cox regression, and random survival forest machine learning algorithm were used to conduct the statistical analysis. Results: Under free breathing, only the maximum POCR differed significantly between standing and prone positions, with prone positions showing higher stress (mean difference = 3.15, p < 0.001). In contrast, the regulated breathing rhythm enhanced performance outcomes compared to free breathing regardless of the starting position. Specifically, exercises initiated from prone positions under regulated breathing recorded a significantly higher maximum and average PHR than those from standing positions (maximum PHR: mean difference = 13.40, p < 0.001; average PHR: mean difference = 6.45, p < 0.001). The multi-variable Cox regression highlighted the starting position as a critical factor influencing the PHR and breathing rhythm as a significant factor for the POCR, with respective variable importances confirmed by the random survival forest analysis. These results underscore the importance of controlled breathing and starting positions in optimizing HIME outcomes. Conclusions: Regulated breathing in high-intensity exercises enhances performance and physiological functions, emphasizing the importance of breathing rhythm over starting position. Effective training should balance exercise volume and technique to optimize performance and minimize stress, reducing overtraining and injury risks.

Hepatic Transcriptome Reveals Potential Key Genes Contributing to Differential Milk Production.

Despite the widespread adoption of TMR or PMR and the formulas designed to sufficiently cover the cows' requirements, individual dairy cows' milk production varies significantly. The liver is one of the most important organs in cow lactation metabolism and plays an essential role in the initiation of lactation. This study aimed to investigate the potential key genes in the liver contributing to the different milk production. We enrolled 64 cows and assigned them to high or low milk yield (MY) groups according to their first 3 weeks of milk production. We performed RNAseq for 35 liver samples with 18 from prepartum and 17 from postpartum cows. The continuous milk yield observation showed a persistently higher milk yield in high MY cows than low MY cows in the first 3 weeks. High MY cows showed better feed conversion efficiency. We identified 795 differentially expressed genes (DGEs) in the liver of high MY cows compared with low MY cows, with up-regulated genes linked to morphogenesis and development pathways. Weighted gene co-expression network analysis (WGCNA) revealed four gene modules positively correlating with milk yield, and protein and lactose yield (p < 0.05). Using the intersected genes between the four gene modules and DEGs, we constructed the linear mixed-effects models and identified six hub genes positively associated and two hub genes negatively associated with milk yield (Coefficients > 0.25, p < 0.05). Random forest machine learning model training based on these eight hub genes could efficiently predict the milk yield (p < 0.001, R2 = 0.946). Interestingly, the expression patterns of these eight hub genes remained remarkably similar before and after parturition. The present study indicated the critical role of liver in milk production. Activated processes involved in morphogenesis and development in liver may contribute to the higher milk production. Eight hub genes identified in this study may provide genetic research materials for dairy cow breeding.

Drought Quantification in Africa Using Remote Sensing, Gaussian Kernel, and Machine Learning

Effective drought management requires precise measurement, but this is challenging due to the variety of drought indices and indicators, each with unique methods and specific uses, and limited ground data availability. This study utilizes remote sensing data from 2001 to 2020 to compute drought indices categorized as meteorological, agricultural, and hydrological. A Gaussian kernel convolves these indices into a denoised, multi-band composite image. Further refinement with a Gaussian kernel enhances a single drought index from each category: Reconnaissance Drought Index (RDI), Soil Moisture Agricultural Drought Index (SMADI), and Streamflow Drought Index (SDI). The enhanced index, encompassing all bands, serves as a predictor for classification and regression tree (CART), support vector machine (SVM), and random forest (RF) machine learning models, further improving the three indices. CART demonstrated the highest accuracy and error minimization across all drought categories, with root mean square error (RMSE) and mean absolute error (MAE) values between 0 and 0.4. RF ranked second, while SVM, though less reliable, achieved values below 0.7. The results show persistent drought in the Sahel, North Africa, and southwestern Africa, with meteorological drought affecting 30% of Africa, agricultural drought affecting 22%, and hydrological drought affecting 21%.

Investigating Spatial Effects through Machine Learning and Leveraging Explainable AI for Child Malnutrition in Pakistan

While socioeconomic gradients in regional health inequalities are firmly established, the synergistic interactions between socioeconomic deprivation and climate vulnerability within convenient proximity and neighbourhood locations with health disparities remain poorly explored and thus require deep understanding within a regional context. Furthermore, disregarding the importance of spatial spillover effects and nonlinear effects of covariates on childhood stunting are inevitable in dealing with an enduring issue of regional health inequalities. The present study aims to investigate the spatial inequalities in childhood stunting at the district level in Pakistan and validate the importance of spatial lag in predicting childhood stunting. Furthermore, it examines the presence of any nonlinear relationships among the selected independent features with childhood stunting. The study utilized data related to socioeconomic features from MICS 2017–2018 and climatic data from Integrated Contextual Analysis. A multi-model approach was employed to address the research questions, which included Ordinary Least Squares Regression (OLS), various Spatial Models, Machine Learning Algorithms and Explainable Artificial Intelligence methods. Firstly, OLS was used to analyse and test the linear relationships among selected variables. Secondly, Spatial Durbin Error Model (SDEM) was used to detect and capture the impact of spatial spillover on childhood stunting. Third, XGBoost and Random Forest machine learning algorithms were employed to examine and validate the importance of the spatial lag component. Finally, EXAI methods such as SHapley were utilized to identify potential nonlinear relationships. The study found a clear pattern of spatial clustering and geographical disparities in childhood stunting, with multidimensional poverty, high climate vulnerability and early marriage worsening childhood stunting. In contrast, low climate vulnerability, high exposure to mass media and high women’s literacy were found to reduce childhood stunting. The use of machine learning algorithms, specifically XGBoost and Random Forest, highlighted the significant role played by the average value in the neighbourhood in predicting childhood stunting in nearby districts, confirming that the spatial spillover effect is not bounded by geographical boundaries. Furthermore, EXAI methods such as partial dependency plot reveal the existence of a nonlinear relationship between multidimensional poverty and childhood stunting. The study’s findings provide valuable insights into the spatial distribution of childhood stunting in Pakistan, emphasizing the importance of considering spatial effects in predicting childhood stunting. Individual and household-level factors such as exposure to mass media and women’s literacy have shown positive implications for childhood stunting. It further provides a justification for the usage of EXAI methods to draw better insights and propose customised intervention policies accordingly.