The increasing popularity of social media in all sectors of society has prompted more companies, including government agencies, to create and manage social media accounts to engage their clients online. For most government agencies, the use of Facebook deals not only with information dissemination but also, with managing reputation, communicating crises, and acting as a communication channel. This paper aims to: a) determine the sentiment of the stakeholders based on their feedback and comments on the NAST PHL Facebook pages from 01 April 2020 to 31 October 2021; b) develop and determine the performance of the different models for sentiment analysis; and c) determine the best model in terms of accuracy, precision, recall, and F1 Score. The researchers used RapidMiner to build a model while the classification process was initiated using Naive Bayes, K-nearest neighbor, and Decision Tree. The performance of each classifier was measured using accuracy, precision, recall, and F1 score. Results showed that the public has a positive sentiment (66 percent) on the various virtual events under the NAST core functions. On the other hand, DT gave the highest values for three out of the four performance measures: precision, recall, and F1 score. However, the values from DT and KNN for accuracy and F1 score were remarkably close, i.e., difference of 0.99 and 0.08, respectively. Based on the results, DT is the best among the three classifiers to use when evaluating the sentiment using Facebook data.