ANALISIS SENTIMEN BERBASIS ASPEK PADA EDOM PEMBELAJARAN MENGGUNAKAN METODE CNN DAN WORD2VEC"

Muhammad, Irfani (2024) ANALISIS SENTIMEN BERBASIS ASPEK PADA EDOM PEMBELAJARAN MENGGUNAKAN METODE CNN DAN WORD2VEC". Undergraduate Thesis thesis, Institut Teknologi Telkom Purwokerto.

	Text cover.pdf Download (433kB)
	Text ABSTRAK.pdf Download (111kB)
	Text ABSTRACT.pdf Download (136kB)
	Text BAB I.pdf Download (118kB)
	Text BAB II.pdf Download (669kB)
	Text BAB III.pdf Download (360kB)
	Text BAB IV.pdf Restricted to Registered users only Download (1MB)
	Text BAB V.pdf Download (79kB)
	Text DAFTAR PUSTAKA.pdf Download (120kB)
	Text LAMPIRAN.pdf Restricted to Registered users only Download (80kB)

Abstract

In order to improve the quality of teaching and guide students towards achieving academic goals. EDOM (Lecturer Evaluation by Students) is an instrument that cannot be ignored in evaluating lecturer performance. By involving students as respondents, EDOM provides valuable feedback regarding their learning experiences, identifies lecturers' strengths and weaknesses, and provides suggestions for improvement. This survey practice has become commonplace in various universities, including ITTP (Telkom Purwokerto Institute of Technology). The EDOM survey combines quantitative questions, which are easier to analyze directly through numerical data, and qualitative questions, which require more in-depth content analysis to understand the sentiment or opinion contained in students' text answers. In an effort to find out more in-depth information regarding student satisfaction, analysis can be carried out using the ABSA (Aspect Bassed Sentiment Analysis) method. ABSA is a computational study of opinions, judgments, attitudes and emotions. Sentiments can be divided into 2 classes, namely posi and negative, while aspects can be broken down into aspects of teaching methods and completeness of material. The data obtained from EDOM ITTP amounted to 5116. After going through the data cleaning process, handling slang words, data labeling, 1960 data were obtained, then to word embedding with word2vec, with modeling using CNN. The results of evaluating the CNN model with the Cunfusion Matrix obtained the best results with an accuracy of 84% on the sentiment dataset, then for the aspect dataset it got an accuracy of 94%. Keywords : deep learning, CNN, word2vec, sentiment analysis, AB

Actions (login required)

View Item