Pengembangan Model Prediksi Readmisi Pasien Berdasarkan Analisis Customer Lifetime Value Menggunakan Metode Ensemble

Shodikin, Mohamad (2024) Pengembangan Model Prediksi Readmisi Pasien Berdasarkan Analisis Customer Lifetime Value Menggunakan Metode Ensemble. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025221043-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (3MB) | Request a copy

Abstract

Saat ini tantangan dalam industri perumahsakitan di Indonesia khususnya dalam hal strategi marketing adalah masih banyaknya rumah sakit yang belum mampu melakukan pemetaan tipologi konsumen berdasarkan tingkat loyalitasnya. Seperti halnya yang terjadi di Rumah Sakit X sebagai RS swasta kelas C di Sidoarjo yang masih belum optimal dalam memetakan tipologi konsumen untuk mengembangkan strategi marketing khususnya dalam hal memprediksi readmisi pasien. Tujuan memprediksi readmisi pasien diharapkan dapat membantu manajer rumah sakit dalam merumuskan kebijakan strategi marketing untuk mempertahankan loyalitas konsumen. Penelitian-penelitian yang menitikberatkan pada pemrosesan histori layanan medis pasien untuk pengembangan model prediksi readmisi pasien berdasarkan analisis Customer Lifetime Value (CLV) menggunakan metode Ensemble masih jarang dilakukan. Berdasarkan latar belakang tersebut di atas, maka penulis meneliti prediksi readmisi pasien berdasarkan analisis Customer Lifetime Value menggunakan metode Ensemble. Nilai CLV membantu perusahaan dan organisasi mengalokasikan sumber daya terbatas yang tersedia untuk pelanggan mereka dengan mengkategorikan dan menetapkan bobot tertentu untuk setiap pelanggan. Pemeringkatan CLV dievaluasi dengan salah satu metode populer yaitu model Recency, Frequency, Monetary dan Interpurchase Time (RFMT) untuk penekanan pada pelanggan yang menguntungkan perusahaan. Hasil penelitian menunjukkan dataset readmisi pasien rawat inap dan rawat jalan dapat dibangun berdasarkan analisis CLV menggunakan nilai RFMT yang berasal dari histori layanan medis dan pembobotan RFMT menggunakan teknik Fuzzy-AHP. Segmentasi pasien rawat inap terdiri dari kategori Champions sebanyak 473 (2%), kategori Loyal Customers sebanyak 1.727 (7%), kategori Potential Loyalist sebanyak 3.874 (16%), kategori Lost Customers sebanyak 18.516 (75%). Segmentasi pasien rawat jalan meliputi kategori Champions sebanyak 3.512 (8%), kategori Loyal Customers sebanyak 5.661 (13%), kategori Potential Loyalist sebanyak 11.070 (25%), kategori Lost Customers sebanyak 23.678 (54%). Kategori Champions, Loyal Customers, dan Potential Loyalist termasuk dalam kelompok pasien readmisi. Sedangkan kategori Lost Customers termasuk dalam kelompok pasien non readmisi. Berdasarkan evaluasi kinerja model prediksi readmisi pasien rawat jalan dan rawat inap, didapatkan bahwa semua metode Ensemble baik model Decision Tree, Gradient Boosting dan Stacking memiliki akurasi 100%. Begitu pula nilai Precision, Recall dan F1-Score pada ketiga model tersebut juga mencapai 100%. Berdasarkan hasil evaluasi kinerja model menunjukkan bahwa metode Ensemble sebagai model prediksi yang dikembangkan dalam penelitian ini dapat mengkategorikan segmen pasien rawat jalan dan rawat inap yang meliputi Champions, Loyal Customers, Potential Loyalist, dan Lost Customers dengan benar.
=================================================================================================================================
Currently, the challenge in the hospital industry in Indonesia, especially in terms of marketing strategy, is that there are still many hospitals that have not been able to map consumer typologies based on their level of loyalty. As is the case at X Hospital, a class C private hospital in Sidoarjo, which is still not optimal in mapping consumer typologies to develop marketing strategies, especially in terms of predicting patient readmissions. The aim of predicting patient readmissions is expected to help hospital managers in formulating marketing strategy policies to maintain consumer loyalty. Research that focuses on processing patient medical service history to develop patient readmission prediction models based on Customer Lifetime Value (CLV) analysis using the Ensemble method is still rarely carried out. Based on the background above, the author examines patient readmission predictions based on Customer Lifetime Value analysis using the Ensemble method. CLV values help companies and organizations allocate the limited resources available to their customers by categorizing and assigning certain weights to each customer. CLV ranking is evaluated using one of the popular methods, namely the Recency, Frequency, Monetary and Interpurchase Time (RFMT) model to emphasize customers who are profitable for the company. The research results show that a readmission dataset for inpatients and outpatients can be built based on CLV analysis using RFMT values derived from medical service history and RFMT weighting using the Fuzzy-AHP technique. The segmentation of inpatients consists of the Champions category with 473 (2%), the Loyal Customers category with 1,727 (7%), the Potential Loyalist category with 3,874 (16%), the Lost Customers category with 18,516 (75%). The outpatient segmentation includes the Champions category with 3,512 (8%), the Loyal Customers category with 5,661 (13%), the Potential Loyalist category with 11,070 (25%), the Lost Customers category with 23,678 (54%). The Champions, Loyal Customers and Potential Loyalist categories are included in the readmission patient group. Meanwhile, the Lost Customers category is included in the group of non-readmission patients.Based on the performance evaluation of outpatient and inpatient readmission prediction models, it was found that all ensemble methods, including Decision Tree, Gradient Boosting and Stacking models, had 100% accuracy. Likewise, the Precision, Recall and F1-Score values for the three models also reach 100%. Based on the results of the model performance evaluation, it shows that the Ensemble method as a prediction model developed in this research can categorize the outpatient and inpatient patient segments which include Champions, Loyal Customers, Potential Loyalists, and Lost Customers correctly.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Readmisi Pasien, Metode Ensemble, Customer Lifetime Value, Patient Readmission, Ensemble Methode
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Mohamad shodikin
Date Deposited:	05 Aug 2024 04:45
Last Modified:	03 Sep 2024 03:21
URI:	http://repository.its.ac.id/id/eprint/113565

Actions (login required)

View Item