SUMANTIAWAN, Dody Indra and Suseno, Jatmiko Endro and Syafei, Wahyul Amien (2023) ANALISIS SENTIMEN ULASAN KONSUMEN MENGGUNAKAN METODE SUPPORT VECTOR MACHINE DAN KOMBINASI SYNTHETIC MINORITY OVERSAMPLING TECHNIQUE (SMOTE) DENGAN TOMEK LINKS. Masters thesis, UNIVERSITAS DIPONEGORO.

	Text 1 COVER.pdf Download (1MB)
	Text 2 BAB I.pdf Download (333kB)
	Text 3 BAB II.pdf Download (530kB)
	Text 4 BAB III.pdf Restricted to Repository staff only Download (593kB)
	Text 5 BAB IV.pdf Restricted to Repository staff only Download (1MB)
	Text 6 BAB V.pdf Download (257kB)
	Text 7 DAFTAR PUSTAKA.pdf Download (332kB)
	Text 8 LAMPIRAN.pdf Restricted to Repository staff only Download (307kB)

Abstract

Aktivitas belanja di pasar online terutama tren fashion dengan segala upaya promo yang ditawarkan terus meningkat. Salah satu pertimbangan untuk membeli produk di pasar online adalah dengan membaca ulasan. Setiap ulasan konsumen menunjukkan tingkat ketertarikan terhadap produk. Banyaknya ulasan negatif dan timbulnya banyak ulasan yang bervariasi menjadi masalah dalam mengkategorikan ulasan. Analisis sentimen sebagai cara melihat polaritas ulasan untuk mengklasifikasi ulasan positif dan negatif. Metode Support Vector Machine dan kombinasi Synthetic Minority Oversampling Technique (SMOTE) dengan Tomek Links diterapkan dalam penelitian ini. Klasifikasi menggunakan metode Support Vector Machine dan kombinasi Synthetic Minority Oversampling Technique (SMOTE) dengan Tomek Links menunjukkan hasil yang lebih baik dengan akurasi 0,92, Precision 0,89, Recall 0,89, dan f1-score 0,89 daripada tanpa kombinasi Synthetic Minority Oversampling Technique (SMOTE) dengan Tomek Links dengan akurasi 0,68, Precision 0,55, Recall 0,99, dan f1-score 0,71 Hasil word cloud sentiment positif menunjukan ulasan ‘bagus’, ‘warna’, ‘barang’, ‘suka’, ‘kirim’, ‘cepat’, ‘langganan’, ‘diskon’, sedangkan ulasan negatif ‘jelek’, ‘celana’, ‘jahit’, ‘ukur’, ‘tipis’, ‘kecewa’, ‘kecil’, ‘pinggang’. Berdasarkan hasil klasifikasi penyeimbangan data dan word cloud dilakukan perbaikan dengan meningkatkan quality control produk secara keseluruhan, menyertakan kebijakan keterlambatan/kerusakan barang pada toko atau deskripsi produk dan meningkatkan kinerja admin dalam melayani pelanggan.
Kata kunci: Analisis Sentimen, Klasifikasi, Support Vector Machine, SMOTE, Tomek Links

Shopping activities in the online market especially fashion trends with all the promo efforts offered continue to increase. One of the considerations for buying products on the online market is to read reviews. Each consumer review shows the level of interest in the product. The number of negative reviews and the emergence of many varied reviews pose a problem in categorizing reviews. Sentiment analysis is a way of looking at the polarity of reviews to classify positive and negative reviews. The Support Vector Machine method and the combination of the Synthetic Minority Oversampling Technique (SMOTE) with Tomek Links are applied in this study. Classification using the Support Vector Machine method and the combination of the Synthetic Minority Oversampling Technique (SMOTE) with Tomek Links showed better results with an accuracy of 0.92, precision of 0.89, recall of 0.89, and f1-score of 0.89 than without combination the Synthetic Minority oversampling Technique (SMOTE) with Tomek Links with an accuracy of 0.68, precision of 0.55, recall of 0.99, and an f1-score of 0.71. Positive sentiment word cloud results show reviews 'bagus', 'warna', 'barang', 'suka', 'kirim', 'cepat', 'langganan', 'diskon', while the reviews are negative 'jelek', 'celana', 'jahit', 'ukur', 'tipis', 'kecewa', 'kecil', 'pinggang'. Based on the results of the classification of balancing data and word clouds, improvements are made by increasing overall product quality control, including policies for delays/damage to goods in stores or product descriptions and improving admin performance in serving customers.
Keywords: Sentiment Analysis, Classification, Support Vector Machine, SMOTE, Tomek Links.

Actions (login required)

View Item