Implementation of Machine Learning Algorithms in ^UNIMA MAPALUS^ Sentiment Analysis
Vivi P Rantung (a*), Djubir R. E. Kembuan (b), Michella Undap (c)

a) Universitas Negeri Manado, Indonesia, b) Universitas Negeri Manado, Indonesia, c) Universitas Negeri Manado, Indonesia

Abstract

Considering the public reflection of Universities is an important thing to evaluate in aiming to improve performances, thus collecting information about these is mandatory. Sources of digital information in this era can provide an overview of the public^s opinions. Manado State University in this case could take advantage of the large amount of text data which available in digital world today to find out current opinions or conditions. The data is taken from various sources such as search engines, online newspapers and social media applications. The data that was successfully crawled with keyword ^UNIMA MAPALUS^ are 1.236 sentences and after going through several stages of text mining it became 14.011 words. This research used two algorithms for word classification : Support Vector Machine (SVM) and Naive Bayes which are divided into three major opinions, namely positive, negative and neutral. Naive Bayes results show the highest accuracy of 93.75%. The classification results are then visualized using a word cloud with detailed vocabulary density: 0.353, readability index: 21.404, average words per sentence: 2511.0.

Keywords: sentiment analysis, Support Vector Machine (SVM), Naive Bayes, Unima Mapalus

Topic: Universal Quality Education