Enhanced multi-lingual Twitter sentiment analysis using hyperparameter tuning k-nearest neighbors

Abstract

Social media is a medium that is often used by someone to express themselves. These various problems on social media have encouraged research in sentiment analysis to become one of the most popular research fields. Various methods are used in sentiment analysis research, ranging from classic machine learning (ML) to deep learning. Researchers nowadays often use deep learning methods in sentiment analysis research because they have advantages in processing large amounts of data and providing high accuracy. However, deep learning also has limitations on the longer computational side due to the complexity of its network architecture. K-nearest neighbor (KNN) is a robust ML method but does not yet provide high-accuracy results in multi-lingual sentiment analysis research, so a hyperparameter tuning KNN approach is proposed. The results showed that using the proposed method, the accuracy level improved to 98.37%, and the classification error (CE) improved to 1.63%. The model performed better than other ML and even deep learning methods. The results of this study indicate that KNN using hyperparameter tuning is a method that contributes to the sentiment analysis classification model using the Twitter dataset.