Published Paper Details:

The available data on social media has contributed to vast research using sentiment analysis. The twitter-based social media represents a gold-mine approach for analyzing the performance of the brand. Large opinions of the people are found over Twitter that are honest, informative, and casual as compared to the formal type of data-survey analysis using magazines or reports. Millions of people share and express their sentiments over the media discussing about the brands whom they interact with. Sentiment analysis has turned out one of the most significant tools in natural language processing because it opens up numerous possibilities to understand people�s sentiments on different topics. The purpose of an aspect-based sentiment analysis is to understand this further and find out what someone is talking about, and whether he likes it or does not like it. There have been various ways to deal with handle this issue, utilizing machine learning. In this thesis, labelled data is used based on polarity and Tweets pre-process and extract unigram features after pre-processing of the tweets. In pre-processing, the noisy data is removed by using tokenization; stop word removal and stemming these processes remove the duplicate data like to repeat words, hash tags and emoji�s. These features and label learn by SVM, ANN and a hybrid of KNN and ANN (hybrid). In proposed experiment Ensemble approach shows improvement in precision and accuracy than other