Pembuatan Model Deteksi Redundansi Data Pasien Untuk Keabsahan Data Dengan Metode Deep Learning

Widodo, Prevandito Wahyu (2024) Pembuatan Model Deteksi Redundansi Data Pasien Untuk Keabsahan Data Dengan Metode Deep Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032222055-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (1MB) | Request a copy

Abstract

Redundansi data pasien di fasilitas kesehatan sering muncul akibat entri ganda, meskipun Standar Operasional Prosedur (SOP) telah diberlakukan. Hal ini mengakibatkan rekam medik menjadi tidak valid dan dapat menyebabkan tindakan medis berikutnya menjadi tidak sesuai. Untuk mengatasi masalah ini, entity matching (EM) dengan pendekatan deep learning (DL) menjadi solusi potensial. Penelitian ini bertujuan meningkatkan keabsahan data pasien dengan mengaplikasikan DL dalam EM, menggunakan model SIF, RNN, Attention, dan Hybrid. Model RNN menunjukkan performa terbaik dengan rata-rata F1-Score (Valid) sebesar 95.59%, Precision sebesar 93.28%, dan Recall sebesar 98.02%, yang kemudian diintegrasikan ke dalam sistem informasi laboratorium berupa fitur dengan nama "Merger Data AI". Pengujian dengan System Usability Scale (SUS) menunjukkan skor rata-rata 70.75, mengindikasikan bahwa pengguna cukup puas dengan fitur ini. Hasil pengujian model dalam menangani redundansi data di beberapa cabang juga menunjukkan tingkat keberhasilan di atas 90%, meskipun ada beberapa kegagalan yang mayoritas disebabkan oleh kolom data yang kosong seperti alamat dan kota. Implementasi model ini diharapkan dapat meningkatkan keakuratan dan efisiensi pengelolaan data pasien di laboratorium klinik Parahita serta memberikan dampak positif pada efisiensi operasional.
===========================================================
Patient data redundancy in healthcare facilities often arises from duplicate entries, despite the implementation of Standard Operating Procedures (SOP). This results in invalid medical records and can lead to inappropriate subsequent medical actions. To address this issue, entity matching (EM) using deep learning (DL) techniques offers a potential solution. This study aims to enhance the validity of patient data by applying DL in EM, utilizing SIF, RNN, Attention, and Hybrid models. The RNN model demonstrated the best performance, with an average F1-Score (Valid) of 95.59%, Precision of 93.28%, and Recall of 98.02%, which was then integrated into the laboratory information system as a feature named "Merger Data AI." Testing with the System Usability Scale (SUS) yielded an average score of 70.75, indicating that users were quite satisfied with this feature. The model's testing in handling data redundancy across several branches also showed a success rate of over 90%, although there were some failures, mostly due to empty data columns such as address and city. The implementation of this model is expected to improve the accuracy and efficiency of patient data management at Parahita Clinical Laboratory and positively impact operational efficiency.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Entity Matching; Deep Learning; Data Redundancy; Data Validity; Entity Matching; Deep Learning; Redundansi Data; Keabsahan Data
Subjects:	T Technology > T Technology (General) > T58.5 Information technology. IT--Auditing
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Prevandito Wahyu Widodo
Date Deposited:	06 Aug 2024 06:03
Last Modified:	06 Aug 2024 06:03
URI:	http://repository.its.ac.id/id/eprint/113418

Actions (login required)

View Item