Search for collections on Undip Repository

ANALISA PENGGUNAAN MODEL HYBRID FEATURE SELECTION DAN JARINGAN SYARAF TIRUAN UNTUK PENYAKIT GINJAL KRONIS

CHOTIMAH, Siti Noor and Warsito, Budi and Surarso, Bayu (2021) ANALISA PENGGUNAAN MODEL HYBRID FEATURE SELECTION DAN JARINGAN SYARAF TIRUAN UNTUK PENYAKIT GINJAL KRONIS. Masters thesis, UNIVERSITAS DIPONEGORO.

[img] Text
LEMBAR PENGESAHAN, DAFTAR ISI DLL_ Siti Noor Chotimah.pdf

Download (963kB)
[img] Text
BAB I_ Siti Noor Chotimah.pdf

Download (204kB)
[img] Text
BAB II_ Siti Noor Chotimah.pdf

Download (508kB)
[img] Text
BAB III_ Siti Noor Chotimah.pdf
Restricted to Repository staff only

Download (662kB)
[img] Text
BAB IV_ Siti Noor Chotimah.pdf
Restricted to Repository staff only

Download (1MB)
[img] Text
BAB V_ Siti Noor Chotimah.pdf
Restricted to Repository staff only

Download (249kB)
[img] Text
DAFTAR PUSTAKA DAN LAMPIRAN_ Siti Noor Chotimah.pdf
Restricted to Repository staff only

Download (944kB)

Abstract

Jumlah pasien penderita penyakit ginjal kronis (PGK) yang terus mengalami peningkatan membuat penyakit ini menjadi salah satu penyakit yang termasuk dalam permasalahan kesehatan publik secara global. Ada banyak faktor yang bisa dikategorikan sebagai diagnosa untuk penyakit ginjal kronis. Penelitian ini bertujuan untuk memilih diagnosa penyakit ginjal kronis yang dianggap memiliki lebih banyak pengaruh diantara diagnosa lainnya. Terdapat tiga tahapan yang dilakukan dalam penelitian ini. Tahapan pertama yaitu memberi peringkat pada fitur diagnosa dari peringkat yang tertinggi hingga terendah dengan menggunakan Information Gain (IG). Tahapan kedua yaitu mengeliminasi fitur yang dianggap memiliki sedikit pengaruh jika dibandingkan dengan fitur lainnya dengan mengguanakan Sequential Backward Feature Selection (SBS). Tahapan ketiga yaitu menggunakan fitur-fitur terpilih sebagai input pada algoritma klasifikasi Jaringan Syaraf Tiruan (JST) Backpropagation. Berdasarkan hasil penelitian, dengan menggunakan hybrid feature selection antara IG dan SBS, terpilih 15 fitur yang dianggap terbaik dari total 18 fitur dengan hasil akurasi sebesar 90% untuk penelitian dengan menggunakan default hidden neuron dan 86% untuk penelitian dengan menggunakan determined hidden neuron.
Kata kunci: diagnosa awal penyakit ginjal kronis, hybrid feature selection, information gain, sequential backward feature selection, jaringan syaraf tiruan

The number of patients with chronic kidney disease (CKD) which continues to increase makes this disease one of the diseases that are included in global public health problems. There are many factors that can be categorized as a diagnosis for chronic kidney disease. This study aims to select the diagnosis of chronic kidney disease that is considered to have more influence among other diagnoses. This study has three steps. The first step is to rank the diagnostic features from the highest to the lowest using Information Gain (IG). The second step is to eliminate features that are considered to have little influence when compared to other features by using Sequential Backward Feature Selection (SBS). The third step is using the selected features as input to the Backpropagation Artificial Neural Network (ANN) classification algorithm. Based on the results of the study, using hybrid feature selection between IG and SBS, 15 features were selected that were considered the best from a total of 18 features with an accuracy of 90% for studies using default hidden neurons and 86% for studies using determined hidden neurons.
Keywords: chronic kidney disease early diagnosis, hybrid feature selection, information gain, sequential backward feature selection, ANN

Item Type: Thesis (Masters)
Uncontrolled Keywords: diagnosa awal penyakit ginjal kronis, hybrid feature selection, information gain, sequential backward feature selection, jaringan syaraf tiruan
Subjects: Sciences and Mathemathic
Divisions: Postgraduate Program > Master Program in Information System
Depositing User: ekana listianawati
Date Deposited: 21 Mar 2023 08:20
Last Modified: 21 Mar 2023 08:20
URI: https://eprints2.undip.ac.id/id/eprint/12096

Actions (login required)

View Item View Item