A technical framework for HR-driven imbalanced classification: pipeline architecture, oversampling algorithms, ablation analysis, and threshold engineering for data scientist job-change prediction

Predictive modelling of employee job-change propensity from HR survey data presents a canonical cluster of machine-learning engineering challenges: structured mixed-type inputs with systematic missingness, moderate class imbalance, and self-reported labels susceptible to noise. This paper presents the technical design and implementation of an end-to-end HR analytics pipeline applied to the HR Analytics: Job Change of Data Scientists dataset ( \(N = 19{,}158\) ; positive rate \(\approx 24.9\%\) ). We formalise the leak-safe imbalanced-learn Pipeline architecture that confines synthetic oversampling strictly to training folds and empirically quantify—via a controlled leaky-versus-clean experiment—that naive pre-split oversampling inflates 5-fold cross-validated AUC by 0.014–0.018 units for SMOTE-family methods, a margin sufficient to reverse configuration rankings. Nine oversampling algorithms are instantiated and evaluated across 90 model \(\times\) sampler configurations. A systematic ablation study of the feature engineering pipeline identifies missingness indicator construction ( \(\Delta \textrm{AUC} = -0.008\) when removed) and out-of-fold smoothed target encoding ( \(\Delta \textrm{AUC} = -0.009\) ) as the two highest-value components. HistGradientBoosting on the unaugmented training distribution achieves the highest holdout ROC-AUC of 0.816 [95% CI 0.800, 0.832]; all pairwise sampler AUC differences fall within this \(\pm 0.016\) estimation band and are not statistically significant. With threshold optimisation applied uniformly across configurations, performance differences between the best unresampled and best oversampled alternatives are within the confidence interval (F1 \(= 0.644\) vs. 0.648; \(\Delta = 0.004\) , overlapping bootstrap CIs), though oversampled configurations achieve meaningfully higher recall (73.6% vs. 78.1%). Generalizable engineering recommendations are distilled for HR classification tasks with high-cardinality categoricals, systematic missingness, and self-reported labels.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-15
DOI
https://doi.org/10.1038/s41598-026-77466-w
Primary Topic
Imbalanced Data Classification Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A technical framework for HR-driven imbalanced classification: pipeline architecture, oversampling algorithms, ablation analysis, and threshold engineering for data scientist job-change prediction

Mehvish Bashir, Md. Nurul Haque Mollah, S.K. Md. Anik Hassan Rabby, Ishtiak Al Mamoon et al.
Scientific Reports
Imbalanced Data Classification Techniques
article

A technical framework for HR-driven imbalanced classification: pipeline architecture, oversampling algorithms, ablation analysis, and threshold engineering for data scientist job-change prediction

Mehvish Bashir, Md. Nurul Haque Mollah, S.K. Md. Anik Hassan Rabby, Ishtiak Al Mamoon, Hriday Chandra Shil, Nimra Khan, Md Luman Jamali, Mir Mohammad Azad, Chantor Btissam, Md. Mashfi Chowdhury Aurin, Anupam Mazumdar, Samrat Ray, Md Foysal Ahmed
article en

Abstract

Predictive modelling of employee job-change propensity from HR survey data presents a canonical cluster of machine-learning engineering challenges: structured mixed-type inputs with systematic missingness, moderate class imbalance, and self-reported labels susceptible to noise. This paper presents the technical design and implementation of an end-to-end HR analytics pipeline applied to the HR Analytics: Job Change of Data Scientists dataset ( \(N = 19{,}158\) ; positive rate \(\approx 24.9\%\) ). We formalise the leak-safe imbalanced-learn Pipeline architecture that confines synthetic oversampling strictly to training folds and empirically quantify—via a controlled leaky-versus-clean experiment—that naive pre-split oversampling inflates 5-fold cross-validated AUC by 0.014–0.018 units for SMOTE-family methods, a margin sufficient to reverse configuration rankings. Nine oversampling algorithms are instantiated and evaluated across 90 model \(\times\) sampler configurations. A systematic ablation study of the feature engineering pipeline identifies missingness indicator construction ( \(\Delta \textrm{AUC} = -0.008\) when removed) and out-of-fold smoothed target encoding ( \(\Delta \textrm{AUC} = -0.009\) ) as the two highest-value components. HistGradientBoosting on the unaugmented training distribution achieves the highest holdout ROC-AUC of 0.816 [95% CI 0.800, 0.832]; all pairwise sampler AUC differences fall within this \(\pm 0.016\) estimation band and are not statistically significant. With threshold optimisation applied uniformly across configurations, performance differences between the best unresampled and best oversampled alternatives are within the confidence interval (F1 \(= 0.644\) vs. 0.648; \(\Delta = 0.004\) , overlapping bootstrap CIs), though oversampled configurations achieve meaningfully higher recall (73.6% vs. 78.1%). Generalizable engineering recommendations are distilled for HR classification tasks with high-cardinality categoricals, systematic missingness, and self-reported labels.

Scientific Reports
Western Illinois University (US), Sher-e-Kashmir University of Agricultural Sciences and Technology of Kashmir (IN), Trine University (US), Uttara University (BD), Comilla University (BD), International University of Business Agriculture and Technology (BD), Arden University (GB), Sidi Mohamed Ben Abdellah University (MA), University of Management and Technology (PK), University of Rajshahi (BD), Jashore University of Science and Technology (BD), Sir Padampat Singhania University (IN), Bangladesh University of Professionals (BD)
Imbalanced Data Classification Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon