Pujadayanti, Irma (2018) Prediksi Rating Otomatis pada Ulasan Produk Kecantikan dengan Metode Naïve Bayes dan N-gram. Sarjana thesis, Universitas Brawijaya.
Abstract
Maraknya produk kecantikan bermunculan juga menggempur Indonesia khususnya produk impor. Hal ini memicu persaingan yang ketat antar pelaku industri produk kecantikan lokal maupun luar negeri. Oleh karena itu, perlu adanya inovasi pada produk mereka. Banyaknya data ulasan dalam berbagai sumber online berguna sebagai bahan kajian bagi pihak produsen untuk melakukan inovasi pada produk mereka. Bagi Konsumen data tersebut berguna sebagai informasi sebelum membeli produk. Data ulasan tak jarang juga belum disertai dengan rating sehingga membuat produsen mengalami kesulitan dalam mengkategorikan ulasan kedalam sentiment tertentu. Pada penelitian ini membantu mempercepat pengkategorian ulasan kedalam sentiment yang berbentuk rating. Sistem yang dibangun pada penelitian ini menggunakan metode klasifikasi naïve bayes dan penambahan metode n-gram pada pre-processing. Penggunaan n-gram meliputi unigram, bigram dan kombinasi unigram dan bigram bertujuan meningkatkan hasil klasifikasi. Pada pengujian sistem hasil terbaik pada skenario full pre-processing pada semua n-gram. Akurasi unigram 50%, 93%, 93% sedangkan akurasi bigram adalah 39%, 87%, 83% dan akurasi tertinggi adalah kombinasi 49%, 97%, 96% dengan model pengujian toleransi 0, toleransi 1 dan sentiment ulasan. Hasil pengujian menunjukkan penggunaan kombinasi n-gram cukup efektif dalam menyelesaikan masalah dalam penelitian.
English Abstract
The rise of beauty products also pound Indonesia especially imported products. This has triggered intense competition between local and foreign beauty products industry players. Therefore, the need for innovation in their products. The large number of review data in various online sources is useful as a review material for producers to innovate their products. For Consumer the data is useful as information before buying the product. The review data is often also has not been accompanied by a rating that makes manufacturers have difficulty in categorizing into a certain sentiment. In this study helps to accelerate the categorization of reviews into sentiment in the form of rating. The system built on this research uses the naïve bayes classification method and the addition of n-gram method to pre-processing. The use of n-grams including unigram, bigram and combination of unigram and bigram aims to improve the classification results. On testing the best result system in full pre-processing scenario on all n-grams. Accuracy of 50%, 93%, 93% unigram while the accuracy of bigram is 39%, 87%, 83% and the highest accuracy is a combination of 49%, 97%, 96% with tolerance 0, tolerance 1 and sentiment reviews. The results showed that the use of n-grams was enough effective in solving the problems in the study.
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | SKR/FTIK/2018/192/051801214 |
| Uncontrolled Keywords: | Prediksi Rating, Ulasan, Naïve Bayes, N-Gram |
| Subjects: | 000 Computer science, information and general works > 005 Computer programming, programs, data |
| Divisions: | Fakultas Ilmu Komputer > Teknik Informatika |
| Depositing User: | Yusuf Dwi N. |
| Date Deposited: | 31 May 2018 06:42 |
| Last Modified: | 27 Oct 2021 03:43 |
| URI: | http://repository.ub.ac.id/id/eprint/11309 |
Preview |
Text
BAB IV.pdf Download (1MB) | Preview |
Preview |
Text
Bagian Depan.pdf Download (751kB) | Preview |
Preview |
Text
BAB VII.pdf Download (312kB) | Preview |
Preview |
Text
BAB V.pdf Download (670kB) | Preview |
Preview |
Text
BAB VI.pdf Download (508kB) | Preview |
Preview |
Text
Daftar Pustaka.pdf Download (321kB) | Preview |
Preview |
Text
Lampiran.pdf Download (438kB) | Preview |
Preview |
Text
BAB I.pdf Download (440kB) | Preview |
Preview |
Text
BAB II.pdf Download (623kB) | Preview |
Preview |
Text
BAB III.pdf Download (487kB) | Preview |
Actions (login required)
 |
View Item |