Explainable Minimal-Feature Machine Learning for Malaria Screening Using Haematological Parameters: Secondary Analysis of a Public Bangladesh Clinical Dataset
Salma Nur Auliyya Raihanah1*, Lilik Djuari2
Abstract
Background: Routine complete blood count (CBC) parameters can support malaria screening in resource-limited areas, but the optimal minimum number of parameters required remains unclear. Methods: We performed a secondary analysis of a public dataset comprising 2,188 patients with malaria-like symptoms from Bangladesh. Using a stratified five-fold cross-validation with random forest feature ranking, we reduced 12 demographic and haematological candidate predictors down to the top four. The optimized model was evaluated on a 20% held-out test set (n = 438). Results: Platelet count, red cell distribution width, eosinophil count, and mean corpuscular haemoglobin concentration were consistently selected across all folds. On the test set, this four-predictor model achieved an area under the receiver operating characteristic curve (AUC) of 0.820 (95% CI: 0.780–0.856), sensitivity of 0.771, and specificity of 0.714. Performance was statistically equivalent to the full 12-predictor model (DeLong’s test, p = 0.695), with a positive net benefit demonstrated across all decision thresholds (0.01–0.89). Shapley additive explanations identified platelet count as the dominant predictor. Four routine haematological parameters may therefore support low-cost malaria triage in resource-limited settings, pending external and prospective validation. Conclusion: A minimalist set of four routine haematological parameters can effectively support low-cost malaria triage in resource-constrained settings, pending prospective external validation.
Keywords
malaria; machine learning; feature selection; explainable artificial intelligence; haematological parameters; clinical decision support
Cite This Article
Raihanah, S. N. A., Djuari, L. (2026). Explainable Minimal-Feature Machine Learning for Malaria Screening Using Haematological Parameters: Secondary Analysis of a Public Bangladesh Clinical Dataset. International Journal of Scientific Advances (IJSCIA), Volume 7| Issue 6: Nov – Dec 2026, Pages 1014-1025. URL: https://www.ijscia.com/wp-content/uploads/2026/11/Volume7-Issue6-Nov-Dec-No.1124-1014-1025.pdf
Volume 7 | Issue 6: Nov – Dec 2026
