Research on the Nonlinear Relationship Between Carbon Emissions from Residential Land and the Built Environment: A Case Study of Susong County, Anhui Province Using the XGBoost-SHAP Model

Land, Journal Year: 2025, Volume and Issue: 14(3), P. 440 - 440

Published: Feb. 20, 2025

Residential land is the basic unit of urban-scale carbon emissions (CEs). Quantifying and predicting CEs from residential are conducive to achieving urban neutrality. This study took 84 communities in Susong County, Anhui Province as its research object, exploring nonlinear relationship between built environment land. By identifying through building electricity consumption, 14 indicators, including area (LA), floor ratio (FAR), greening (GA), density (BD), gross (GFA), use mix rate (Phh), permanent population (PPD), were selected establish an interpretable machine learning (ML) model based on XGBoost-SHAP attribution analysis framework. The results show that, first, goodness fit XGBoost reached 91.9%, prediction accuracy was better than that gradient boosting decision tree (GBDT), random forest (RF), Adaboost model, traditional logistic model. Second, compared with other ML models, explained influencing factors more clearly. SHAP indicate BD, FAR, Phh most important affecting CEs. Third, there a significant threshold effect characteristic variables Fourth, interaction different dimensions environmental factors, played dominant role interaction. Reducing FAR considered be effective CE reduction strategy. provides practical suggestions for planners reducing land, which has policy implications significance.

Language: Английский

---

Prediction of moderate to severe bleeding risk in pediatric immune thrombocytopenia using machine learning

European Journal of Pediatrics, Journal Year: 2025, Volume and Issue: 184(5)

Published: April 7, 2025

Language: Английский

---

Development and validation of inpatient mortality prediction models for patients with hyperglycemic crisis using machine learning approaches

BMC Endocrine Disorders, Journal Year: 2025, Volume and Issue: 25(1)

Published: March 27, 2025

Hyperglycemic crisis is one of the most common and severe complications diabetes mellitus, associated with a high motarlity rate. Emergency admissions due to hyperglycemic remain prevalent challenging. This study aimed develop validate predictive models for in-hospital mortality risk among patients admitted emergency department using various machine learning (ML) methods. A multi-center retrospective was conducted across six large general adult hospitals in Chongqing, western China. Patients diagnosed were identified an electronic medical record (EMR) database. Demographics, comorbidities, clinical characteristics, laboratory results, complications, therapeutic interventions extracted from records construct prognostic prediction model. Seven algorithms, including support vector machines (SVM), random forest (RF), recursive partitioning regression trees (RPART), extreme gradient boosting dart booster (XGBoost), multivariate adaptive splines (MARS), neural network (NNET), boost (AdaBoost) compared logistic (LR) predicting crisis. Stratified sampling used split data into training (80%) validation (20%) sets. Ten-fold cross performed on set optimize model hyperparameters. The sensitivity, specificity, positive negative values, area under curve (AUC) accuracy all computed comparative analysis. total 1668 eligible present study. rate 7.3% (121/1668). In set, feature importance scores calculated each eight models, top 10 significant features identified. demonstrated good capability, areas value exceeding 0.9 F1 score between 0.632 0.81, except MARS Six algorithm outperformed referred Among selected RPART, RF, SVM achieved best performance (AUC values 0.970, 0.968 0.968, 0.652, 0.762, 0.762 respectively). Feature analysis novel predictors mechanical ventilation, age, Charlson Comorbidity Index, blood gas index, first 24-hour insulin dosage, fluid intake. Most algorithms exhibited excellent model, RPART These overlapping but different, up predictors. Early identification high-risk these could decision-making potentially improve prognosis patients. Not applicable.

Language: Английский

---

“First, Do No Harm” in the Digital Era: Examining the Practicality of the European Health Data Space Proposal and Ethical Implications of Artificial Intelligence. A Systematic Literature Review

Published: Jan. 1, 2025

Language: Английский

---