Perbandingan Evaluasi Kernel Support Vector Machine Untuk Klasifikasi Sentimen Apple Vision Pro Pada Sosial Media X

This study aims to analyze sentiment related to Apple Vision Pro using the Support Vector Machine (SVM) algorithm and evaluate the performance of various kernels, including Linear, Polynomial, Radial Basis Function (RBF), and Sigmoid. The data was collected from platform X through a crawling process, resulting in 1,825 tweets categorized as positive, negative, or neutral, with automatic labeling using VADER. Prior to analysis, the data underwent preprocessing steps such as case folding, data cleaning, stopword removal, lemmatization, and tokenization. The results indicate that the Linear and Polynomial kernels achieved an accuracy of 81% with a 70:30 and 80:20 data split using the TF-IDF technique without SMOTE. However, applying SMOTE to the Polynomial kernel significantly improved accuracy to 99%, highlighting the importance of addressing data imbalance to enhance model performance. This study also suggests exploring other algorithms, such as Random Forest, Gradient Boosting, or Deep Learning models (LSTM, BERT), as well as more meaningful word representation techniques like Word2Vec or GloVe, to improve sentiment analysis accuracy in future research. These findings provide valuable insights into the effectiveness of SVM in selecting the optimal kernel for sentiment classification.