Analysing text for emotions like happiness or sadness is called sentiment analysis. In order to narrow down a large feature collection, this work employs a number of feature selection methods, including Mutual information, Chi-square, Information gain, and TF-if. These procedures are assessed using a dataset of 2000 MOVIE reviews. Support vector machine from the weka9 tool is used to carry out the categorization. We also look at the question of what feature works best for determining reviewers' emotions. Word functions, POS tags, and larger word structures are also part of our feature set.