Rofiqoh, Umi (2017) Analisis Sentimen Tingkat Kepuasan Pengguna Penyedia Layanan Telekomunikasi Seluler Indonesia pada Twitter dengan Metode Support Vector Machine dan Lexicon Based Features. Sarjana thesis, Universitas Brawijaya.
Abstract
Analisis sentimen adalah salah satu cabang penelitian dari Text Mining yang berguna untuk mengklasifikasi dokumen teks berupa opini berdasarkan sentimen. Dokumen teks yang digunakan dalam penelitian berasal Twitter tentang opini masyarakat mengenai penyedia layanan telekomunikasi seluler. Metode yang digunakan adalah Support Vector Machine dengan menggunakan Lexicon Based Features sebagai pembaharuan fiturnya selain memakai fitur TF-IDF. Data yang digunakan pada penelitian ini sebanyak 300 data yang dibagi menjadi dua jenis data dengan perbandingan 70% untuk data latih dan 30% untuk data uji. Hasil akurasi sistem yang diperoleh dari analisis sentimen dengan metode Support Vector Machine dan Lexicon Based Features sebesar 79% menggunakan nilai degree sebesar 2, nilai konstanta learning rate 0,0001, serta jumlah iterasi maksimum sebanyak 50 kali. Sedangkan sistem analisis sentimen tanpa menggunakan Lexicon Based Features menghasilkan tingkat akurasi sebesar 84% dengan nilai parameter yang sama.
English Abstract
Sentiment Analysis Is A Part Of Research From Text Mining Which Is Usefull To Classify Text Documents Contained Opinion Based On Sentiment. Text Document That Is Used In Research Comes From Twitter From People’s Opinion About Cellular Telecommunication Service Provider. The Used Method Is Support Vector Machine With Using Lexicon Based Features As Its Feature Renewal Instead Of Using TF-IDF Features. The Used Data In This Research Is 300 Data Which Divided Into Two Types Of Data With Ratio 70% For Training Data And 30% For Testing Data. The Result Of System Accuracy That Is Obtained From Sentiment Analysis Using Support Vector Machine And Lexicon Based Features Method Is 79% Using Degree Value 2, Constant Learning Rate Value 0.0001, And Maximum Iteration Is 50 Times. While Sentiment Analysis System Without Using Lexicon Based Features Is Resulting Accuracy At 84% With The Same Parameter Values.
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | SKR/FTIK/2017/347/051706389 |
| Uncontrolled Keywords: | analisis sentimen, Twitter, Support Vector Machine, Lexicon Based Features |
| Subjects: | 000 Computer science, information and general works > 004 Computer science > 004.6 Interfacing and communications |
| Divisions: | Fakultas Ilmu Komputer > Teknik Informatika |
| Depositing User: | Kustati |
| Date Deposited: | 23 Aug 2017 02:35 |
| Last Modified: | 13 Dec 2020 11:12 |
| URI: | http://repository.ub.ac.id/id/eprint/1557 |
Preview |
Text
Umi Rofiqoh.pdf Download (7MB) | Preview |
Actions (login required)
 |
View Item |