Article Per Year (5 Year)
p-Index From 2021 - 2026
0.562
P-Index
This Author published in this journals
All Journal JISKa (Jurnal Informatika Sunan Kalijaga)
Maria Ulfah Siregar
Sunan Kalijaga State Islamic University Yogyakarta
Computer Science & IT Electrical & Electronics Engineering Library & Information Science

Published : 3 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 3 Documents
Search

 1 
Optimasi Hyperparameter Ensemble Learning untuk Prediksi Perkembangan Penyakit Diabetes dengan Explainable AI David Suharjanto; Muhammad Syafiq Akmal; Nur Fikri Khuluq; Muh Naufal Muzhaffar; Maria Ulfah Siregar
JISKA (Jurnal Informatika Sunan Kalijaga) Vol. 11 No. 2 (2026): May 2026
Publisher : UIN Sunan Kalijaga Yogyakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14421/jiska.5953

Abstract

This research focuses on optimizing and assessing ensemble learning models for predicting diabetes progression by combining hyperparameter tuning and explainable artificial intelligence techniques. Experiments were conducted using the scikit-learn diabetes dataset, which contains 442 samples with ten numerical features representing patients’ clinical conditions. The data were split into 80% for training and 20% for testing. Two ensemble methods were explored: Random Forest Regressor (bagging) and XGBoost Regressor (boosting). Hyperparameter optimization was carried out using RandomizedSearchCV and BayesianSearchCV under a five-fold cross-validation scheme. Model performance was evaluated using MAE, MSE, RMSE, and R² metrics, while interpretability was examined through SHAP summary plots. The results indicate that BayesianSearchCV consistently delivered superior performance gains compared to random search. In particular, the optimized XGBoost model achieved an R² score of 0.5018, improving by 19.8% over the baseline model (R² = 0.4188), and reduced RMSE from 55.49 to 51.37. SHAP analysis showed that serum triglycerides, body mass index, and blood pressure were the most influential features. Overall, the findings suggest that Bayesian-based hyperparameter optimization can effectively improve ensemble regression performance in medical prediction tasks involving limited datasets.
Optimizing Iris Plant Classification with Ensemble Models and XAI: A Comprehensive Analysis of Model Performance Ahmad Subadri; Ishmah Afiyah; Fiki Sanora; Arya Indrawan; Maria Ulfah Siregar
JISKA (Jurnal Informatika Sunan Kalijaga) Vol. 11 No. 2 (2026): May 2026
Publisher : UIN Sunan Kalijaga Yogyakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14421/jiska.5967

Abstract

This study aims to improve the performance of Iris plant classification by integrating ensemble learning techniques with Explainable Artificial Intelligence (XAI) to achieve high accuracy while enhancing model interpretability. Random Forest, XGBoost, and AdaBoost algorithms are combined within a Voting Ensemble framework and evaluated using the Iris Plants dataset, which comprises 150 data samples distributed equally across three Iris species (50 samples per class: Iris setosa, Iris versicolor, and Iris virginica). The dataset exhibits a perfectly balanced class distribution, ensuring that no class imbalance correction was required. The Voting Ensemble model was evaluated using a hold-out test set (80:20 split) and further validated through 5-Fold Stratified Cross-Validation, yielding a mean cross-validation accuracy of 95.83% (±2.64%) and a test set accuracy of 93.33%. To enhance model transparency, the SHAP (SHapley Additive Explanations) method is applied to explain the contribution of each feature to the prediction outcomes. The Voting Ensemble model achieved an ROC AUC score of 0.9900 (macro-average), with Precision, Recall, and F1-Score each reaching 0.9333 (macro-average). Feature importance analysis reveals that petal length and petal width are the primary factors in the Iris species classification process. The strong correlation (r = 0.9991) between feature importance scores in the Random Forest model and SHAP values confirms the consistency and reliability of the model’s interpretability. These findings demonstrate that integrating ensemble learning with XAI not only improves predictive performance but also strengthens transparency and trust in machine learning models, particularly for plant classification tasks.
Fine-Tuned Artificial Neural Network (ANN) with Explainable AI (XAI) for Classification and Interpretation of the Breast Cancer Wisconsin Ikra Muhibban Ananta; Hanif Amarudin; Restu Adi Akbar; M. Fajratul Ikhsan; Maria Ulfah Siregar
JISKA (Jurnal Informatika Sunan Kalijaga) Vol. 11 No. 3 (2026): September 2026
Publisher : UIN Sunan Kalijaga Yogyakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14421/jiska.5973

Abstract

Breast cancer remains one of the leading causes of cancer-related mortality among women, making early detection and accurate diagnosis essential. Although machine learning and deep learning approaches have achieved high classification performance, their interpretability and evaluation robustness remain limited. This study proposes a Fine-Tuned Artificial Neural Network (ANN) integrated with Explainable Artificial Intelligence (XAI) for breast cancer classification using the Wisconsin Diagnostic Breast Cancer (WDBC) dataset. Model evaluation was conducted using Stratified 5-Fold Cross-Validation, while optimization was performed through learning rate adjustment and early stopping. Model interpretability was analyzed using SHapley Additive exPlanations (SHAP), including summary, decision, force, and waterfall plots. Experimental results show that the baseline ANN achieved an average accuracy of 97.54%, precision of 97.82%, recall of 98.32%, and F1-score of 98.04%. The Fine-Tuned ANN improved performance, achieving 97.89% accuracy, 98.08% precision, 98.60% recall, and 98.32% F1-score, with better stability across folds. SHAP analysis identified worst area, worst radius, worst texture, and mean concave points as the most influential features. Furthermore, the ablation study indicated that the Sigmoid activation function, Adam optimizer, and learning rate of 0.01 produced the optimal configuration.
Co-Authors Ahmad Subadri Arya Indrawan David Suharjanto Fiki Sanora Hanif Amarudin Ikra Muhibban Ananta Ishmah Afiyah M. Fajratul Ikhsan Muh Naufal Muzhaffar Muhammad Syafiq Akmal Nur Fikri Khuluq Restu Adi Akbar