Social media platforms, such as Facebook, Twitter, etc., are very popular ways to express the opinions of a person(s) and are used to interact with different people in the online world. A collection of tweets can provide a reflection of public sentiment towards various events. Unfortunately, in the wrong hands, these platforms can be used to spread malicious contents. Spams in Twitter have become a very critical problem. So in this paper, we find out a positive or negative sentiment on tweets by using a well-known machine learning method for text categorization. Additionally, we manually label (positive/negative) tweets to build a trained method to accomplish the task. The task is to look for a correlation between twitter sentiment and events that have occurred. The trained model is based on the Naive Bayes classification method and Different Algorithms to Predict the Accuracy Percentage.Our aim is to detect the sentiment and the spam or fake accounts and Compare with Differe