The last few months, Indonesia and the world are being hit by a disaster caused by the covid-19 virus, which resulted in a global pandemic that has infected more than one million human beings, because of this many countries have begun to implement various policies, including Indonesia, which is currently is implementing social distancing and PSBB, the application of which raises the pros and cons of people from all fields, to find out the response and response of the community related to the policy, a survey is needed to analyze the positive and negative comments, which are on Twitter, to analyze the response of the community can be done with text mining, text mining is the process of finding the meaning of an unstructured text or writing, to analyze the response of the community data taken from social media Twitter in the form of comments / tweets with a total of 400 comments and divided into 200 positive sentiment comments in 200 negative sentiment comments, the data is processed using the SVM classification method and will be compared with the SVM classification by PSO feature selection, the text preprocessing used is tokenize, transform cases, stopword filter and generate n-grams, and for the word weighting process using tf-idf , the accuracy of the SVM classification without PSO is 67.00% with the AUC value obtained is 0.774 and the accuracy of the SVM classification with PSO is 98.25% with the AUC value of 0.999, from the results the SVM with PSO feature selection is better in classifying sentiments than SVM without PSO feature selection.