Indonesia Decision Tree Outperforms Naive Bayes For Infidelity Divorce Sentiment Classification

Parabelem Tinno Dolf Rompas, Shella Sasoeng, Gladly Caren RorimpandeyNon-Scopus19 Juni 2026Afiliasi: ID

Abstrak

General Background Divorce due to infidelity has become a massive public issue widely exposed on social media platforms. Specific Background This surge in digital opinion generates unstructured textual data reflecting polarized moral perceptions, necessitating machine learning to map public sentiment. Knowledge Gap Previous sentiment analysis predominantly focused on public services, leaving a significant void in examining domestic conflicts like marital betrayal. Aims This study analyzes public perception on social media X and compares Naïve Bayes and Decision Tree algorithms on an imbalanced dataset. Results Processing 3,450 tweets using TF-IDF and SMOTE revealed that public opinion is dominantly Neutral (47.9%) and Negative (46.7%), with Positive sentiment at only 5.3%. The Decision Tree algorithm achieved 96.95% accuracy, significantly outperforming Naïve Bayes (73.59%). Novelty This research proves the Decision Tree model's superior stability and accuracy over Naïve Bayes in handling imbalanced textual data regarding sensitive moral issues. Implications The developed model and its interactive dashboard offer a robust foundation for content moderation and formulating contextual family resilience education programs. Highlights The rule-based algorithm achieved 96.95% accuracy, demonstrating exceptional stability when analyzing textual data. Public reactions on the X platform regarding marital betrayal are primarily divided into neutral and negative categories. Implementing the Synthetic Minority Over-sampling Technique effectively addressed dataset imbalances prior to model evaluation. Keywords Decision Tree; Naive Bayes; Sentiment Analysis; Social Media X; Synthetic Minority Oversampling Technique

Identifier

Ekspor BibTeX
Scroll to Top