Article Per Year (5 Year)
p-Index From 2021 - 2026
0.408
P-Index
This Author published in this journals
All Journal Indonesian Journal of Electronics, Electromedical Engineering, and Medical Informatics
Andi Farmadi
Department of Computer Science, Faculty of Mathematics and Natural Science, Lambung Mangkurat University, Banjarbaru, Indonesia
Computer Science & IT Control & Systems Engineering Decision Sciences, Operations Research & Management Electrical & Electronics Engineering Health Professions Materials Science & Nanotechnology

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

 1 
Comparison Between K-Fold Cross Validation And Percentage Split In Decision Tree Algorithms For Anemia Classification Nanda Putri Rahmawati; Irwan Budiman; Muhammad Itqan Mazdadi; Andi Farmadi; Friska Abadi
Indonesian Journal of Electronics, Electromedical Engineering, and Medical Informatics Vol. 8 No. 1 (2026): February
Publisher : Jurusan Teknik Elektromedik, Politeknik Kesehatan Kemenkes Surabaya, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35882/ijeeemi.v8i1.315

Abstract

Anemia is a significant global health challenge characterized by a pathological deficit in hemoglobin concentration, often leading to physiological instability. Accurate clinical diagnosis typically relies on complete blood count (CBC) tests, which provide critical hematological parameters for classification. While machine learning models have demonstrated high efficacy in diagnosing anemia, existing research often relies on static data partitioning strategies that may overlook evaluation reliability and performance stability. This study addresses this gap by shifting the focus from architectural benchmarking to validation robustness, specifically evaluating the C4.5 algorithm's performance across different data-splitting techniques. The research uses a dataset comprising 1,281 clinical records with 14 numerical features and 9 anemia-type labels. To assess stability, two distinct partitioning strategies were implemented: a static Percentage Split (ranging from 60:40 to 90:10) and iterative K-Fold Cross Validation (with K values of 3, 5, 7, 10, and 15). Experimental results demonstrate that the C4.5 algorithm achieved its peak performance with the 90:10 Percentage Split, achieving an average accuracy of 99.46%, precision of 98.32%, and recall of 99.28%. In comparison, the K-Fold (K=10) approach yielded a slightly lower but more stable accuracy of 99.19% with a significantly reduced standard deviation (±0.09), highlighting its reliability for clinical applications. While the high-ratio percentage split maximizes training exposure and predictive potential, the K-Fold method provides a more objective, generalizable benchmark by accounting for the entire data distribution. The study further identifies challenges in classifying minority classes, such as Leukemia with thrombocytopenia, due to inherent data scarcity. Ultimately, this research confirms that the C4.5 algorithm, when paired with an optimal partitioning protocol, remains a robust and highly interpretable solution for clinical anemia screening, outperforming several complex modern architectures
Comparative Evaluation of TabKANet with Oversampling and Feature Selection Ablation for Software Defect Prediction Muhammad Faza Azhiman Saputra; Setyo Wahyu Saputro; Mohammad Reza Faisal; Radityo Adi Nugroho; Andi Farmadi
Indonesian Journal of Electronics, Electromedical Engineering, and Medical Informatics Vol. 8 No. 3 (2026): August
Publisher : Jurusan Teknik Elektromedik, Politeknik Kesehatan Kemenkes Surabaya, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35882/ijeeemi.v8i3.351

Abstract

Software defect prediction (SDP) focuses limited testing resources on the modules most likely to fail, but real-world software metric data are tabular, noisy, and severely class-imbalanced, which degrades conventional learners. The Kolmogorov-Arnold Network (KAN) and Transformer architectures recently achieved strong results on tabular data, yet their combined form, TabKANet, has not been evaluated for SDP, nor has the contribution of common preprocessing techniques been quantified. This study adapts and comparatively evaluates TabKANet against established baselines and measures the contribution of oversampling and feature selection through a structured ablation. Twelve all-numerical NASA Metrics Data Program datasets were used. The pipeline applied duplicate removal, MinMax normalization, effective class weighting, and stratified five-fold cross-validation, with oversampling (SMOTE) and Recursive Feature Elimination (RFE) inserted inside the training folds. Four TabKANet variants (A: base, B: +SMOTE, C: +RFE, D: +SMOTE+RFE) were compared with Multi-Layer Perceptron (MLP), standalone KAN, and TabNet, and differences were tested with the Wilcoxon signed-rank test at a 0.05 significance level. The base TabKANet (variant A) achieved the highest mean AUC of 0.7603, slightly ahead of MLP (0.7594) and KAN (0.7583) and well above TabNet (0.7092). Its advantage over TabNet was significant (p = 0.002), whereas it was statistically equivalent to MLP and KAN (p = 0.733). TabNet attained the highest recall (0.739) but the lowest precision (0.228), indicating over-prediction of defects, while TabKANet kept precision and recall balanced. In the ablation, SMOTE significantly reduced AUC (p = 0.042), RFE caused no significant change (p = 0.733), and their combination stayed neutral (p = 0.266). TabKANet therefore performed best without additional resampling. TabKANet is thus a competitive architecture for all-numerical, highly imbalanced SDP, matching strong neural baselines and surpassing TabNet, where effective class weighting alone suffices and SMOTE is counter-productive.
Co-Authors Friska Abadi Irwan Budiman Mohammad Reza Faisal Muhammad Faza Azhiman Saputra Muhammad Itqan Mazdadi Nanda Putri Rahmawati Radityo Adi Nugroho Setyo Wahyu Saputro