2026-10-14T15:54:22Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33989088 2026-09-24T17:45:17Z category_734 category_106 category_122 portal_5 item_type_3 month_year_09_2026

<p>Performance of four classification models (XGBoost, Gradient Boosting, Logistic Regression, and Random Forest) in predicting depression within each latent cluster. Values represent the mean area under the ROC curve (AUC) along with 95% confidence intervals (CI) computed on held-out test sets. Models were trained using the top 30 cluster-specific features identified by permutation feature importance (PFI), with class imbalance addressed via SMOTE. XGBoost consistently achieved the highest AUC across all clusters.</p> Divya Sharma (1751311) Alice Rueda (15399721) Qiaowei Lin (8982809) Shakila Meshkat (12360623) Argyrios Perivolaris (21799471) Venkat Bhat (11814479) Biological Sciences not elsewhere classified Science Policy Mental Health standardized mean differences multifactorial disorder influenced machine learning classifiers explainable deep learning existing studies tend 804 &# 8211 latent health phenotypes extensive health surveys large analytic cohort aware screening research precision mental health discovering depression subgroups top 30 pfi permutation feature importance nutrition examination survey derived multimodal data national health precision psychiatry pfi ). oriented research multimodal data analytic sample 30 train depression within depression risk variational autoencoder transparent prediction trained within thereby furthering test split specifically demographic social determinants ranked features questionnaire data integrating high identify integrated five modalities enhance interpretability driven patterns domains independently clustering within 9 %. 874 ). <p>Performance of four classification models (XGBoost, Gradient Boosting, Logistic Regression, and Random Forest) in predicting depression within each latent cluster. Values represent the mean area under the ROC curve (AUC) along with 95% confidence intervals (CI) computed on held-out test sets. Models were trained using the top 30 cluster-specific features identified by permutation feature importance (PFI), with class imbalance addressed via SMOTE. XGBoost consistently achieved the highest AUC across all clusters.</p> 2026-09-24T17:45:05Z Dataset Dataset 10.1371/journal.pdig.0001719.t002 https://figshare.com/articles/dataset/_p_Performance_of_four_classification_models_XGBoost_Gradient_Boosting_Logistic_Regression_and_Random_Forest_in_predicting_depression_within_each_latent_cluster_Values_represent_the_mean_area_under_the_ROC_curve_AUC_along_with_95_confidence/33989088 CC BY 4.0