Skripsi
ANALISIS SENTIMEN MASYARAKAT TERHADAP IBU KOTA NUSANTARA MENGGUNAKAN METODE ENSEMBLE SOFT VOTING
The Ibu Kota Nusantara (IKN) project has sparked various opinions on social media. This research aims to develop a web-based sentiment analysis system using the Ensemble Soft Voting method. The Multinomial Naive Bayes algorithm was selected for its efficiency in probability-based text classification, while Support Vector Machine (SVM) is reliable in handling high-dimensional data. This combination aims to balance the unique characteristics of both algorithms to achieve prediction stability. Utilizing 4,372 clean opinions and TF-IDF weighting, the system was developed using the Rational Unified Process (RUP) methodology. The 5-Fold Cross-Validation testing yielded optimal parameters of Alpha=1.0 (Naive Bayes) and a Linear Kernel with C=1.0 (SVM). In the final testing, the single SVM model excelled in identifying negative sentiments (90% Specificity). Overall, however, the Ensemble Soft Voting method proved to be the most stable, achieving an Accuracy, Precision, and F1-Score of 90%, along with an 89% Specificity. In conclusion, this method effectively integrates the probability strengths of Naive Bayes and the margin sharpness of SVM in predicting public opinions regarding IKN.