PENGARUH KERNEL POLYNOMIAL TERHADAP DATA NON-LINEAR DALAM IDENTIFIKASI HATESPEECH BERBAHASA INDONESIA PADA KOMENTAR INSTAGRAM MENGGUNAKAN METODE SUPPORT VECTOR MACHINE

Puspitaningrum HP, Farisa Yumna (2023) PENGARUH KERNEL POLYNOMIAL TERHADAP DATA NON-LINEAR DALAM IDENTIFIKASI HATESPEECH BERBAHASA INDONESIA PADA KOMENTAR INSTAGRAM MENGGUNAKAN METODE SUPPORT VECTOR MACHINE. Other thesis, UPN "Veteran" Yogyajarta.

	Text ABSTRAK.pdf Download (206kB)
	Text COVER.pdf Download (276kB)
	Text DAFTAR ISI.pdf Download (411kB)
	Text DAFTAR PUSTAKA.pdf Download (447kB)
	Text PENGESAHAN PEMBIMBING.pdf Download (213kB)
	Text PENGESAHAN PENGUJI.pdf Download (179kB)
	Text SKRIPSI FULL_FARISA YUMNA PUSPITANINGRUM HP.pdf Restricted to Repository staff only Download (3MB)

Abstract

vi
ABSTRAK
Support Vector Machine (SVM) merupakan metode pembelajaran mesin supervised
learning yang telah terbukti sebagai salah satu algoritma pembelajaran yang paling kuat
untuk kategorisasi teks, tetapi karena pada prinsipnya bekerja secara linear sehingga untuk
mengatasi masalah data non-linear dikembangkan fungsi kernel. Terdapat beberapa kernel
non-linear salah satunya seperti kernel polynomial. Kernel polynomial memiliki parameter
degree yang dapat disesuaikan sehingga mendapatkan hasil yang optimal. Penelitian ini
melakukan analisis pengaruh kernel polynomial terhadap data non-linear yang
diimplementasikan dalam identifikasi hatespeech berbahasa Indonesia pada komentar
Instagram. Untuk mengetahui data yang dipakai berbentuk linear atau non-linear, maka
digunakan teknik reduksi Principal Component Analysis (PCA) untuk memvisualisasikan
data. Kernel yang digunakan yaitu kernel polynomial, linear, dan RBF.
Hasil pengujian menunjukkan bahwa metode Support Vector Machine
menggunakan kernel polynomial memiliki performa yang lebih baik dibandingan kernel
linear dan kernel RBF dalam melakukan identifikasi data non-linear berjumlah 918
komentar yang terdiri dari 463 non-hatespeech dan 455 hatespeech dengan proses split
data training dan testing 80:20, serta melalui tujuh proses preprocessing yaitu case folding,
cleaning, remove repetation character, tokenizing, normalisasi, stemming, dan stopword
removal. Selain itu, penentuan nilai parameter C, degree, dan gamma juga memiliki
pengaruh dalam meningkatkan performa pada model. Performa terbaik kernel polynomial
diperoleh dengan nilai parameter C=0.1, degree=1, dan gamma=0.01 yaitu akurasi
83.15%, presisi 83.45% dan recall sebesar 83.15%. Sedangakan kernel RBF diperoleh
performa terbaik pada nilai parameter C=2 dan gamma=0.001 yaitu akurasi 80.43%, presisi
80.51%, recall 80.43%, dan kernel linear performa terbaik didapatkan dengan nilai
parameter C=0.01 yaitu akurasi, presisi, dan recall sebesar 77.17%.
Kata kunci: analisis sentimen, ujaran kebencian, Instagram, data non-linear, kernel
polynomial, Support Vector Machine, Principal Component Analysis
vii
ABSTRACT
Support Vector Machine (SVM) is a supervised learning machine learning method
that has proven to be one of the most powerful learning algorithms for text categorization,
but because it works linearly in principle so to overcome the problem of non-linear data, a
kernel function was developed. There are several non-linear kernels, one of which is the
polynomial kernel. The polynomial kernel has degree parameters that can be adjusted to
get optimal results. This research analyzes the effect of the polynomial kernel on non-
linear data implemented in the identification of Indonesian-language hatespeech on
Instagram comments. To find out the data used is linear or non-linear, the Principal
Component Analysis (PCA) reduction technique is used to visualize the data. The kernels
used are polynomial, linear, and RBF kernels.
The test results show that the Support Vector Machine method using the polynomial
kernel has better performance than the linear kernel and RBF kernel in identifying non-
linear data totaling 918 comments consisting of 463 non-hatespeech and 455 hatespeech
with a split training and testing data process of 80:20, and through seven preprocessing
processes namely case folding, cleaning, removing repetation characters, tokenizing,
normalizing, stemming, and stopword removal. In addition, determining the value of the
parameters C, degree, and gamma also has an influence in improving the performance of
the model. The best performance of the polynomial kernel is obtained with a parameter
value of C = 0.1, degree = 1, and gamma = 0.01, namely 83.15% accuracy, 83.45%
precision and 83.15% recall. While the RBF kernel obtained the best performance at
parameter value C = 2 and gamma = 0.001, namely accuracy 80.43%, precision 80.51%,
recall 80.43%, and linear kernel the best performance is obtained with parameter value C
= 0.01, namely accuracy, precision, and recall of 77.17%.
Keywords: sentiment analysis, hatespeech, Instagram, non-linear data, kernel polynomial,
Support Vector Machine, Principal Component Analysis

Item Type:	Thesis (Other)
Subjects:	Z Bibliography. Library Science. Information Resources > ZA Information resources
Divisions:	Faculty of Engineering, Science and Mathematics > School of Engineering Sciences
Depositing User:	Eko Yuli
Date Deposited:	09 Oct 2023 03:11
Last Modified:	09 Oct 2023 03:11
URI:	http://eprints.upnyk.ac.id/id/eprint/37908

Actions (login required)

View Item