Analisis Dan Prediksi Biaya Layanan Bpjs Berdasarkan Tindakan Penanganan Penyakit Menggunakan Ensemble Learning

Nommensen, Ingwer Ludwig (2024) Analisis Dan Prediksi Biaya Layanan Bpjs Berdasarkan Tindakan Penanganan Penyakit Menggunakan Ensemble Learning. Other thesis, Institut Teknologi Sepuluh November.

Text 5025201259-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (4MB) | Request a copy

Abstract

Rumah sakit di Indonesia mencapai 3000 unit, BPJS mengalami kerugian 5 trilliun rupiah (2014-2015), dan diperkirakan defisit 11 trilliun rupiah pada Agustus-September 2025. Untuk meminimalisir kerugian, prediksi biaya BPJS menggunakan Ensemble Learning dibuat karena lebih efektif menangani data bervariasi dibandingkan Single Learning dan membutuhkan daya komputasi lebih rendah dibandingkan Deep Learning. Penelitian ini menggunakan Ensemble Learning dengan dataset BPJS (773.288 baris, 56 kolom) untuk mengurangi kerugian BPJS. Data dipraproses terlebih dahulu, kemudian dibagi menjadi Data Training dan Data Testing (rasio 4:1), dan diuji dengan berbagai teknik regresi seperti CatBoost, Random Forest, Decision Tree, Linear, Lasso Regression, Support Vector Regression, dan Single Layer FeedForward Neural Network (SLFN). Model berbasis kelompok penyakit dipilih karena lebih efektif daripada model clustering atau non-clustering, mengingat INACBG adalah pembayaran berbasis CaseMix. Berdasarkan pengujian, rata-rata R2 Score untuk 12 kelompok penyakit adalah 0,84775, dengan rata-rata MAPE Score sebesar 0,10516. Decision Tree Regression tanpa Parameter Tuning pada kelompok Ambulatory Groups Episodic menjadi model terbaik dengan R2 Score 0,931 dan MAPE 0,032. Meskipun harga layanan bervariasi dan data kelompok penyakit tidak seimbang, Ensemble Learning terbukti lebih unggul dibandingkan Single Learning dan Extreme Learning Machine, menjadi model dengan MAPE terbaik di 9 dari 14 kelompok penyakit.
=================================================================================================================================
Hospitals in Indonesia reached 3000 units, BPJS suffered a loss of 5 trillion rupiah (2014-2015), and an estimated deficit of 11 trillion rupiah in August-September 2025. To minimize losses, BPJS cost prediction using Ensemble Learning is made because it is more effective in handling variable data than Single Learning and requires lower computing power than Deep Learning. This research uses Ensemble Learning with BPJS dataset (773,288 rows, 56 columns) to reduce BPJS losses. The data is preprocessed, then divided into Training Data and Testing Data (4:1 ratio), and tested with various regression techniques such as CatBoost, Random Forest, Decision Tree, Linear, Lasso Regression, Support Vector Regression, and Single Layer FeedForward Neural Network (SLFN). Disease group-based models were chosen because they are more effective than clustering or non-clustering models, given that INACBG is a CaseMixbased payment. Based on the test, the average R2 Score for 12 disease groups was 0.84775, with an average MAPE Score of 0.10516. Decision Tree Regression without Parameter Tuning on Ambulatory Groups Episodic was the best model with R2 Score 0.931 and MAPE 0.032. Despite variable service prices and unbalanced disease group data, Ensemble Learning proved superior to Single Learning and Extreme Learning Machine, being the model with the best MAPE in 9 out of 14 disease groups.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Regresi, Penyakit, Prediksi Biaya, BPJS, Regression, Disease, Cost Prediction
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Ingwer Ludwig Nommensen
Date Deposited:	08 Feb 2024 09:07
Last Modified:	08 Feb 2024 09:07
URI:	http://repository.its.ac.id/id/eprint/106623

Actions (login required)

View Item