Adaptive Synthetic Support Vector Machine Multiclass untuk mengklasifikasikan Imbalance data pada Sentimen kenaikan Bahan Bakar Minyak

Abstract

The phenomenon of increasing fuel oil (BBM) has become a trending topic for people in Indonesia in September 2022. Indonesian people from various provinces have chosen their opinions regarding this phenomenon on social media, one of which is Twitter. Sentiment analysis is used to describe a person's opinion on social media about a phenomenon. In this study, we will look at public sentiment regarding the increase in fuel prices which is labeled into three categories, namely positive, neutral and negative. This study applies adaptive synthetics to overcome data imbalances caused by negative sentiment. The data used in this research is public opinion related to the increase in fuel prices in every province in Indonesia. Each province is limited to 100 opinions. The classification method applied to this research is the multiclass Support Vector Machine (SVM). The results obtained are that people in all provinces in Indonesia have a negative opinion regarding the increase in fuel prices. The results of the multiclass SVM classification show an average accuracy of 87.94%, with the highest accuracy of 95%.