<p>Nowadays people tend to express opinions and their emotions about a wide range of subjects across social networks. As a result, where millions of users engage with one another via social networks generate enormous amounts of data. The emotions of the users can be identified with the analysis of this data. Networks like Twitter represent the public’s aggregate wisdom and opinion, making them useful for getting immediate input to identify emotions of the users. The primary objective of the proposed work is to do sentiment analysis on the textual data. Understanding the precise mood within a given category is frequently largely dependent on the intensity or quantity of feelings and thoughts. The proposed model follows three phases viz., Data Collection, exploration and pre-processing phase, negation handling, extracting useful features and sentiment calculation phase, and model building for classification and prediction phase. In the first phase, the raw data is transformed to useful information for further processing. Negation words have substantial influence in identifying sentiment polarity. Non-handling of negation words leads to misclassification. Negation words are handled in second phase. Finally, for emotion detection Logistic Regression, Naïve Bayes, K-Nearest Neighbor and Support Vector Machine classification models are used. The novelty of proposed model lies in identification and replacement of negation words with their antonyms to accurately classify the emotion. The results have shown that Logistic Regressor along with the proposed algorithm has given better results in detecting and classifying emotions with an accuracy of 88.75% when compared with Logistic Regressor alone.</p>