Klasifikasi Penyakit Gagal Jantung Mengunakan Metode Decision Tree

Abstract

Deteksi awal gagal jantung sangat krusial untuk mengurangi angka sakit dan kematian. Metode machine learning, khususnya klasifikasi yang berbasis decision tree, menunjukkan potensi untuk mendukung keputusan medis dengan memisahkan pasien berisiko menggunakan variabel klinis yang biasa. Tujuan penelitian ini adalah untuk merancang dan menilai model Decision Tree dalam mengklasifikasikan pasien dengan gagal jantung menggunakan data klinis yang bersifat publik. Langkah-langkah dalam penelitian mencakup preprocessing (mengatasi nilai yang hilang, normalisasi, dan pemilihan fitur), pelatihan dengan stratified k-fold cross-validation, serta penilaian menggunakan metrik accuracy, precision, recall, F1-score, dan AUC. Hasil dari eksperimen menunjukkan bahwa Decision Tree yang dioptimalkan memberikan performa yang kompetitif serta keunggulan dalam interpretabilitas melalui aturan keputusan yang jelas. Sumbangan penelitian ini meliputi (1) pipeline yang dapat direproduksi untuk klasifikasi gagal jantung (heart failure), (2) kumpulan aturan yang mendukung skrining klinis heuristik, dan (3) perbandingan empiris terhadap metode machine learning lainnya. Temuan ini menunjukkan bahwa Decision Tree dapat menjadi alat skrining awal yang efektif, terutama di tempat dengan keterbatasan sumber daya.Spotting heart failure (HF) early is super essential for reducing illness and death. Methods using machine learning, especially tree-based methods for sorting, are looking promising for helping doctors make choices by identifying which patients are at risk using standard clinical info. This research is about creating and testing a Decision Tree model to classify heart failure patients using a clinical dataset accessible to anyone. We use careful prep work (dealing with missing info, making things the right size, and picking out what’s important), followed by teaching the model online by using a clever way of speeding up the data and making sure our performance measures are solid. We evaluate the model’s performance using metrics such as accuracy, precision, recall, F1-score, and the area under the ROC curve (AUC). Our tests show that a Decision Tree that’s been tweaked just right does an outstanding job of sorting things and has the bonus of being easy to understand—which is key for doctors to actuauseecause it makes clear rules that doctors can look at. The research also looks at how the Decision Tree did compare to what other machine learning studies have found on heart data recently, pointing out the compromises between being simple and easy to understand versus how well it predicts in more complex or hard-to-understand models. What this paper adds to the table using a Decision Tree with data that’s free for anyone to use, (1) a set of rules that makes sense and can be turned into simple checks for doctors to use, and (3) a real-world comparison to basic ML ways that have been talked about in other papers. The results suggest that Decision Tree models could be a helpful and easy-to-understand tool for doing early checks for HF when there aren’t a lot of resources.