Robust Rare Event Detection with TAMS–DCSL in Noisy Imbalanced Tabular Data: Composite Severe Anthropometric Failure

Read full text Summarise with VIRAT
Authors: Joko Minardi, Fikri Budiman, Ahmad Zainul Fanani, Nova Rijati
Cited by: 0
References: 20

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Routine growth-monitoring records enable scalable screening of severe child undernutrition, yet learning reliable models is difficult because severe cases are rare and routine measurements introduce feature/label noise.We propose TAMS-DCSL, a two-stage framework for rare-event classification in noisy, highly imbalanced tabular data.Topology-Aware Minority Sampling (TAMS) generates synthetic minority instances while preserving local neighborhood structure and restricting sampling to safe-core minority regions identified via neighborhood density and label consistency, reducing noise and class overlap.Dynamic Cost-Sensitive Learning (DCSL) updates misclassification costs during training using validation feedback, mitigating the brittleness of static cost ratios.We evaluate on a child growth monitoring dataset (N=47,964) with a composite severe anthropometric failure label (SAF) based on WHO severe z-score criteria.Stratified k-fold results improve minority recall from 0.643 to 0.785 and GMean from 0.801 to 0.863 versus standard XGBoost while maintaining high specificity.Public benchmarks confirm robustness under extreme imbalance and noisy label conditions.

Fields of Study & Subjects

Imbalanced Data Classification Techniques Machine Learning and Data Classification Data Stream Mining Techniques

Publication details

Journal / source
International journal of intelligent engineering and systems
Volume
19
Issue
4
Pages
303–313
Publication date
23 March 2026
Language
EN