Sentiment analysis or opinion mining is proven to be very effective to analyze huge and complex amount of text of social media. Social media provides an online environment for the users to show their behaviors and emotions through tweets and post. Massive amount of personal information is placed on the World Wide Web due to huge usage of social media. Moods and emotions of the user are different from each other. Sentiment analysis of any written text especially social media content is applicable to extract the opinions, emotions and meaningful insights for better decision making. There are many challenges in the accurate and reliable sentiment analysis of available social media content. The challenges can be both technical and theoretical. Machine learning-based sentiment analysis techniques have issues such as huge lexicon, semantic gap, handling of negation, domain dependency and bi-polar words. Previously, many machine learning and data mining techniques have been proposed by several researchers to resolve these issues. However, the existing techniques have failed to provide satisfactory and reliable results for most of the available datasets. A novel methodology is proposed to overcome above mentioned issues using better and simplified way with less computational complexity and high reliability. Data acquisition, feature encoding, data preprocessing, feature selection, and classification are the various phases of implemented framework. Data gathering and preprocessing step is very critical in the analysis of data. The proposed research mainly contributes during data preprocessing, feature encoding, and classification phases. In feature encoding phase, a hybrid approach of bi-gram and tri-gram is used for word embedding. In the experiments, several benchmark datasets have been utilized to evaluate the effectiveness of the proposed framework. The results obtained from the proposed xii methodology show better or at least comparable results with maximum confidence. The outcome of the proposed work will be helpful to enhance the process of sentiment analysis of social media contents. The experimental results of the framework will be validated using WEKA simulation software.