Classification Sentiment Toward the Indonesian National Soccer Team on Twitter Using Text Mining Transformation

Abstract

People now primarily use social media, particularly Twitter, to share their thoughts, feelings, and reactions to events, including in sports like soccer.  By gathering information from the official Twitter account @TimnasIndonesia during the World Cup qualifying phase, this study seeks to examine how the public views the Indonesian national team.  The Support Vector Machine (SVM) approach was used to classify 412 tweets after they had undergone text pre-processing steps such as data cleaning and text transformation.  Three sentiment categories were employed: good, negative, and neutral. With a percentage of 76.7%, neutral sentiment is the most prevalent sentiment, followed by positive sentiment (17.0%) and negative sentiment (6.3%), according to the classification results.  With a precision of 0.83 and a recall of 1.00, the neutral category outperformed the others, according to the model evaluation.  The model's overall accuracy rate of 83% indicates how successful the strategy is.  Still, there are issues with categorizing positive and negative emotions.  There are still a lot of positive tweets that go undetected since positive emotion has a very low recall (0.18) and a high precision (1.00). Thus, it is advised that future studies concentrate on developing more representative text features and enhancing the classification performance of minority categories using methods like oversampling, undersampling, or class weight adjustment. This will help to balance the data distribution and enable the model to classify all sentiment categories more accurately.