ANALISIS SENTIMEN DAMPAK COVID-19 TERHADAP EKONOMI INDONESIA PADATEXT TWITTER MENGGUNAKAN METODE STOCHASTIC GRADIENT DESCENT DAN SELEKSI FITUR QUERY EXPANSION RANKING

Saaban, Nisfu (2023) ANALISIS SENTIMEN DAMPAK COVID-19 TERHADAP EKONOMI INDONESIA PADATEXT TWITTER MENGGUNAKAN METODE STOCHASTIC GRADIENT DESCENT DAN SELEKSI FITUR QUERY EXPANSION RANKING. Other thesis, UPN "Veteran" Yogyajarta.

	Text ABSTRAK.pdf Download (155kB)
	Text COVER.pdf Download (209kB)
	Text DAFTAR ISI.pdf Download (158kB)
	Text DAFTAR PUSTAKA.pdf Download (138kB)
	Text LEMBAR PENGESAHAN PEMBIMBING.pdf Download (595kB)
	Text SKRIPSI FULL_NISFU SAABAN_123170018.pdf Restricted to Repository staff only Download (6MB)

Abstract

vii
ABSTRAK
Penelitian analisis sentimen dampak covid-19 terhadap ekonomi indonesia
menerapkan metode stochastic gradient descent atau SGD untuk mengolah fitur yang
banyak pada data text twitter namun dalam proses training metode SGD bersifat sensitif
terhadap fitur yang diberikan sehingga mempengaruhi hasil akurasi dan prediksi.
Masalah SGD sensitif terhadap fitur dapat diatasi dengan menggunakan metode
seleksi fitur QER, dari penelitian yang dilakukan diterapkan langkah untuk memberikan
normalisasi bobot pada tiap fitur di tiap kelas positif dan negatif, berdasarkan bobot yang
diberikan selanjutnya digunakan perhitungan score pada fitur dari hasil normalisasi bobot
tersebut dan fitur akan diurutkan berdasarkan score tertinggi dengan tujuan untuk
msssengetahui fitur yang berpengaruh sehingga SGD dapat langsung mengambil fitur yang
berpengaruh pada model.
Berdasarkan hasil penelitian metode SGD tampa QER tidak mampu untuk mengenali
fitur yang diberikan sehingga akurasi hanya mendapatkan 63%, hasil prediksi yang
ditunjukan juga tidak sesuai dengan kelas asli dari fitur, sedangkan untuk SGD dengan QER
mampu mengenali fitur dan hasil akurasi meningkat menjadi 75%.
Kata Kunci : Analisis sentimen, stochastic gradient descen, query expansion
ranking, klasifikasi
viii
ABSTRACT
Sentiment analysis research on the impact of Covid-19 on the Indonesian economy
applies the stochastic gradient descent or SGD method to process a lot of features in Twitter
text data but in the training process the SGD method is sensitive to the features provided so
that it affects accuracy and prediction results.
The SGD problem which is sensitive to features can be overcome by using the QER
feature selection method. From the research carried out, steps are applied to provide
normalized weights for each feature in each positive and negative class, based on the weights
given, then the score calculation is used for the features from the results of normalizing these
weights and the features will be sorted based on the highest score with the aim of knowing
the influential features so that the SGD can immediately take the features that affect the
model.
Based on the research results the SGD method without QER is unable to recognize
the features provided so that accuracy only gets 63%, the prediction results shown are also
not in accordance with the original class of features, while for SGD with QER it is able to
recognize features and the resulting accuracy increases to 75%.
Keywords: Sentiment analysis, stochastic gradient descent, query expansion ranking,
classification

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Sentiment analysis, stochastic gradient descent, query expansion ranking, classification
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:	08 Jun 2023 02:03
Last Modified:	08 Jun 2023 02:03
URI:	http://eprints.upnyk.ac.id/id/eprint/35834

Actions (login required)

View Item