COMPARISON OF SUPPORT VECTOR MACHINE BASED ON FASTTEXT WITHOUT AND WITH FIREFLY OPTIMIZATION PARAMETERS FOR DISASTER SENTIMENT ANALYSIS IN INDONESIA

Abstract

Sentiment analysis is a process for analyzing opinions, sentiments, assessments, and emotions from someone's statements regarding a domain or is also a process for entering and processing data in the form of text. Support vector machine (SVM) is a supervised machine learning technique that functions as a separator of two classes of data. SVM aims to obtain numerical vectors using fasttext. SVM cannot choose appropriate parameters so the use of parameters is not optimal. To obtain optimal parameters with better classification results, firefly optimization was carried out. This research compares the fasttext-based SVM method without and with firefly optimization parameters using data from tweets with the keyword "Indonesian disaster" which was crawled using the Twitter application. The results of this research obtained 128 dimensions that form the weight of each word. This means that each word is represented in a 128-dimensional vector space. The evaluation of the SVM classification model with and without firefly optimization provides an accuracy of 89.1% and 61.3% respectively. This shows that the SVM classification method with firefly optimization provides quite good classification performance compared to the SVM model without optimization.