AULIA, Hozana and Wibowo, Adi and Sutrisno, Sutrisno (2025) INTEGRASI RANDOM FOREST, ADASYN, DAN SHAP UNTUK PREDIKSI DIABETES DAN INTERPRETABILITAS. Masters thesis, UNIVERSITAS DIPONEGORO.
|
Text
Cover-1.pdf Download (118kB) |
|
|
Text
Cover.pdf Restricted to Repository staff only Download (1MB) |
|
|
Text
Bab I.pdf Download (355kB) |
|
|
Text
Bab II.pdf Download (694kB) |
|
|
Text
Bab III.pdf Restricted to Repository staff only Download (480kB) |
|
|
Text
Bab IV.pdf Restricted to Repository staff only Download (989kB) |
|
|
Text
Bab V.pdf Restricted to Repository staff only Download (343kB) |
|
|
Text
Daftar Pustaka.pdf Download (325kB) |
|
|
Text
Lampiran.pdf Restricted to Repository staff only Download (409kB) |
Abstract
Diabetes merupakan salah satu penyakit kronis dengan prevalensi yang tinggi secara global dan berpotensi menyebabkan komplikasi serius jika tidak dideteksi sejak dini. Salah satu tantangan dalam pengembangan model prediksi diabetes adalah ketidakseimbangan data serta kurangnya interpretasi dari hasil prediksi yang dihasilkan oleh model berbasis machine learning. Penelitian ini bertujuan untuk mengembangkan integrasi metode random forest, ADASYN, dan SHAP sebagai pendekatan terpadu dalam membangun model klasifikasi diabetes yang tidak hanya akurat, tetapi juga dapat dijelaskan secara lebih transparan. Dataset yang digunakan pada penelitian ini adalah PIMA Indians, dimana dataset ini tidak seimbang. Metode penelitian mencakup preprocessing data, penyeimbangan kelas menggunakan ADASYN, pelatihan model menggunakan random forest, serta interpretasi hasil prediksi dengan SHAP. Hasil penelitian menunjukkan bahwa model menghasilkan akurasi sebesar 82%, precision 70%, recall 83% dan F1-score 76%. Fitur-fitur Glucose, BMI, dan Age ditemukan sebagai kontributor utama terhadap prediksi. Analisis SHAP dilanjutkan dengan mengidentifikasi bahwa individu dengan usia 30-50 tahun dan obesitas merupakan kelompok berisiko tinggi. Model yang dibangun memberikan kontribusi terhadap sistem informasi prediktif berbasis machine learning yang tidak hanya efektif, tetapi juga dapat dipertanggungjawabkan secara interpretatif untuk mendukung keputusan klinis.
Kata Kunci: Prediksi Diabetes, Random Forest, ADASYN, SHAP, Interpretabilitas, Machine Learning
Diabetes is one of the chronic diseases with a high prevalence globally and has the potential to cause serious complications if not detected early. One of the challenges in developing a diabetes prediction model is the imbalance of data and the lack of interpretation of the prediction results generated by machine learning-based models. This research aims to develop the integration of random forest, ADASYN, and SHAP methods as an integrated approach in building a diabetes classification model that is not only accurate, but also more transparently explainable. The dataset used in this study is PIMA Indians, where this dataset is unbalanced. The research methods include data preprocessing, class balancing using ADASYN, model training using random forest, and interpretation of prediction results with SHAP. The results showed that the model produced an accuracy of 82%, precision 70%, recall of 81% and an F1-score of 75%. Features Glucose, BMI, and Age were found to be major contributors to the prediction. The SHAP analysis was followed by identifying that individuals aged 30-50 years and obesity were a high-risk group. The model built contributes to a predictive information system based on machine learning that is not only effective, but also interpretively accountable to support clinical decisions.
Keywords: Diabetes Prediction, Random Forest, ADASYN, SHAP, Interpretability, Machine Learning
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Prediksi Diabetes, Random Forest, ADASYN, SHAP, Interpretabilitas, Machine Learning |
| Subjects: | Sciences and Mathemathic |
| Divisions: | Postgraduate Program > Master Program in Information System |
| Depositing User: | ekana listianawati |
| Date Deposited: | 10 Dec 2025 08:21 |
| Last Modified: | 10 Dec 2025 08:21 |
| URI: | https://eprints2.undip.ac.id/id/eprint/42034 |
Actions (login required)
![]() |
View Item |
