MODEL KLASIFIKASI PERILAKU PENCEGAHAN KANKER SERVIKS BERBASIS MACHINE LEARNING PADA WANITA USIA SUBUR MASYARAKAT PERKOTAAN INDONESIA

sobar, sobar (2016) MODEL KLASIFIKASI PERILAKU PENCEGAHAN KANKER SERVIKS BERBASIS MACHINE LEARNING PADA WANITA USIA SUBUR MASYARAKAT PERKOTAAN INDONESIA. Doctoral thesis, Universitas Andalas.

Preview	Text (cover dan disertasi) COVER DAN ABSTRAK.pdf - Published Version Download (204kB) | Preview
Preview	Text (BAB I) BAB 1.pdf - Accepted Version Download (87kB) | Preview
Preview	Text (BAB VII) BAB VII.pdf - Published Version Download (91kB) | Preview
Preview	Text (DAFTAR PUSTAKA) DAFTAR PUSTAKA.pdf - Published Version Download (141kB) | Preview
	Text (DISERTASI FULL) DISERTASI FULL SOBAR.pdf - Published Version Restricted to Repository staff only Download (5MB)

Abstract

ABSTRAK Nama : S o b a r Program Study : Pascasarjana S3 Ilmu Kesehatan Masyarakat Title : Model Klasifikasi Perilaku Pencegahan Kanker Serviks Berbasis Machine Learning pada Wanita Usia Subur Masyarakat Perkotaan Indonesia Kanker serviks menduduki urutan ketiga dari kejadian kanker pada wanita di seluruh dunia. Terdapat lebih dari 80% kasusnya terjadi di negara berkembang, terutama terjadi pada Wanita Usia Subur (WUS) yang tinggal di perkotaan. Tujuan dari penelitian ini adalah menemukan model klasifikasi perilaku pencegahan kanker serviks pada WUS masyarakat perkotaan di Indonesia berdasarkan variabel-variabel terkait perilaku antara lain: dukungan sosial, pemberdayaan, sikap, norma subjektif, persepsi, motivasi dan niat. Jenis penelitian kuantitatif dengan desain penelitian menggunakan comparative cross sectional study. Populasi penelitian adalah WUS yang terkena dan tidak terkena kanker serviks pada masyarakat perkotaan Indonesia, direpresentasikan di Rumah Sakit Cipto Mangunkusumo dan tinggal di Jakarta. Jumlah sampel sebanyak 106 responden, dimana 35 responden yang terkena kanker serviks dan 71 responden yang tidak terkena kanker serviks. Penelitian dilaksanakan bulan Januari- Februari 2016. Penelitian ini terdiri atas tiga fase analisis; 1). Uji model menggunakan pendekatan Structural Equation Model (SEM), 2). Uji model klasifikasi menggunakan Machine Learning (ML), dan 3). Uji gabungan antara seleksi atribut berbasis SEM dengan uji klasifikasi berbasis ML. Fase analisis pertama memperoleh hasil bahwa perilaku pencegahan kanker serviks ditentukan secara langsung dan tidak langsung oleh variabel dukungan sosial, pemberdayaan, sikap, norma subjektif, persepsi,motivasi dan niat, dimana R-Square (R2) sebesar 59.18%.Variabel norma subyektif berpengaruh langsung paling dominan dengant-statistik2.24dan f2 effect size 0.14, disusul variabel sikap dengan t-statistik 2.88 dan f2 effect size 0.13, dan paling rendah veriabel persepsi dengan t-statistik1.69 danf2 effect size 0.02. Variabel dukungan sosial dan pemberdayaan berpengaruh secara tidak langsung terhadap perilaku pencegahan kanker serviks. Fase analisis kedua, menyebutkan bahwa model pengukuran ini memiliki kinerja (performance) akurasi, sensitivity dan specificity di atas 90% dan Area Under Curve (AUC) di atas 0.95. Dari tujuh algoritma machine learning yang memiliki performance terbaik adalah LR (Logistic Regresion) dengan akurasi sebesar 96.27%, dan AUC sebesar 0.99 diikuti NB (Naïve Bayes) dengan akurasi 96,18% dan AUC sebesar 0,99. Sensitivity dan specificity antara LR dan NB nilainya sama sebesar 94.29% dan 97.18%. Fase analsis ketiga, menyebutkan bahwa terdapat kecenderungan kenaikan kinerja (performance) atau minimal sama antar sebelum dan sesudah seleksi atribut. Berdasarkan hasil analisis tersebut, dapat disimpulkan bahwa model pengukuran klasifikasi perilaku pencegahan kanker serviks berbasis ML merupakan model yang sesuai (fit) dan mampu mengklasifikasikan perilaku pencegahan kanker serviks secara akurat pada WUS masyarakat perkotaan Indonesia. Pencegahan kanker serviks berbasis perilakuadalah menjanjikan (promising) sebagai upaya penting pada level pencegahan primer dan upaya pre-emtifdalam konteks pencegahan suatu penyakit. Kata Kunci: perilaku pencegahan, kanker serviks, dukungan sosial, pemberdayaan, sikap, norma subjektif, persepsi, motivasi, niat. ABSTRACT Nama : S o b a r Programe Study : Pascasarjana S3 Public Health Science Title : Classification Model for Prevention Behavior of Cervical Cancer based on Machine Learning on Urban Women of Childbearing Age in Indonesia Cervical cancer was ranked third of occurrence of cancer in women in the world. There are more than 80% the case occurring in developing countries, especially happens to women of chilebearing age (WCA) in urban community.The aim of this research is to find classifications model for prevention of cervical cancerin urban women of chilebearing age in Indonesia based on related behavior variables, such as: social support, empowerment, attitude, subjective norm, perception, motivation, intention. The kind of research is quantitative with the design the research uses comparative cross sectional study. The population research is WCA that exposed and not exposed to cervical cancer of the urban population in Indonesia, represented at the Cipto Mangunkusumo Hospital and living in Jakarta. The number of samples from 106 respondents, where 35 respondent’s cervical cancer and 71 respondents were not affected cervical cancer. Research held in January to February 2016. Research is composed of the three phases analysis; 1) A model test used Structural Equation Model (SEM), 2)The model classifications test used Machine Learning (ML) and 3)Test a combination of SEM based selection attribute with ML based classification. In fisrt phase analysis, the result shows that preventionbehavior cervical cancer is determined by direct and indirect by those sevent variables with R-Square (R2) is 59.18%. Attitude variable directly influence most dominant with t-statistic 2.88 and f2 effect size 0.13, followed by subjective norms variable with t-statistic 2.24and f2 effect size 0.02. Social support and empowerment variables are influential indirectly prevention behavior cervical cancer. IN second phase analysis, the classification model has performance accuracy above 90 % and Area Under Curve (AUC) on the 0.95. There are seven algorithm machine learning; Naïve Bayes (NB), Neural Network (NN), Decision Tree (DT),Logistic Regresion (LR), Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), and k-Nearest Neighboar (kNN), withthe best is LR with accuracy of 96.27 % and AUC of 0.99 folowed by NB with accuracy of 96,18% and AUC of 0,99. In third phase analysis,the results of the accuracy of being increased after conducted selection attribute based on SEM result with the best is LR 96.36 % in accuracy and 0,996 in AUC. According to the analysis, it can be concluded that the model classification measurement behavior cervical cancer prevention is the appropriate (fit) and able to classify behavior cervical cancer prevention accurately in urban WCA in Indonesia. Cervical cancer prevention based behavior an important aspect at the level of prevention primary and promise (promising) as an effort to pre-emtif in the context of disease prevention. Keywords: prevention behavior, cervical cancer, social support, empowerment, attitude, subjective norm, perception, motivation, intentions.

Actions (login required)

View Item