Bayesian Network Based Sentiment Analysis of Political Statements

In order to solve important issues including managing unbalanced datasets, capturing complex connections among linguistic characteristics, and guaranteeing model interpretability, this study proposes a Bayesian Network-based method for sentiment analysis of political utterances. Existing models frequently have issues with inadequate transparency, handling class imbalance, and inefficiencies in probabilistic reasoning, despite recent breakthroughs in sentiment analysis employing deep learning and ensemble techniques. The absence of efficient feature selection and the requirement for improved class balance strategies are further difficulties. This work provides an improved Bayesian Network model that incorporates cutting-edge feature engineering methods including group-based statistics, interaction terms, and polynomial transformations in order to address these problems. Probabilistic inference is utilized for precise sentiment categorization, and the Synthetic Minority Oversampling Technique (SMOTE) is applied to rectify class imbalances. The proposed model accomplishes an accuracy of 57 percent and a recall of 57 percent, with precision remaining an area for future improvement. Compared to conventional machine learning models, this approach offers superior scalability, interpretability, and adaptability, making it a robust solution for nuanced sentiment analysis in political discourse.