PERBANDINGAN METODE-METODE KLASIFIKASI UNTUK ANALISIS SENTIMEN BERDASARKAN PEMBENTUKAN TEKNIK VECTOR SPACE MODEL
Daftar Isi:
- The main task in sentiment analysis is to classify sentiments. There are several classification methods based on the type of classification approach, including the probability approach to the appearance of an attribute in a class, one of which is the Multinomial Naïve Bayes; approach to the distance of an attribute, one of which is Support Vector Machine; and binary decision approach to an attribute, one of them is Logistic Regression. Before classifying sentiments, a technique for transforming sentiments in the form of text into numerical techniques is called a vector space model technique. Vector space model techniques include Bag-Of-Words and Doc2Vec. This study will compare the application of the Multinomial Naïve Bayes classification method, Support Vector Machine and Logistic Regression to the Bag-Of-Words and Doc2Vec vector space model techniques. The most effective sentiment classification results in Bag Of Words vector space model technique is using the Multinomial Naïve Bayes classification method with an average accuracy of 85.140% and an average classification time of 0.154 seconds. In the vector space model of the Doc2Vec model, the most effective sentiment classification results use the Logistic Regression classification method with an average accuracy of 62.467% and an average classification time of 0.450 seconds.