Twitter, a social media platform with millions of users, serves as a valuable source for unique insights. The case of Lestibillar domestic violence has garnered attention, fueling various circulating rumors that encompass positive, negative, and neutral opinions. This, in turn, gives rise to the potential spread of fake news. To counter this, sentiment analysis is employed using machine learning techniques. In this research, two machine learning algorithms within the realm of supervised learning are compared: lexicon-based and Naive Bayes. Sentiment objects are created for each algorithm to facilitate the comparison, aiming to determine which algorithm performs better in terms of accuracy. The results of the calculations indicate that Naive Bayes outperforms, achieving a superior accuracy of 99.96%, while the lexicon-based method lags significantly behind at 10.29%. The dominance of positive tweets is evident, comprising 2709 out of the total tweets on Twitter.