The growth of user?s generated contents increased exponentially onthe microblogging platforms like Facebook, Twitter and Blogger in the form of comments and opinions. This bulk of helpful data is difficult to analyse and also a time consuming task. So,an intelligent text mining system that can automatically analyse and categorise such vast user generated data is needed. Due to the noisiness in data, it is difficult to design such a processing technique because of spelling mistakes, grammatical errorsand improper punctuation. Automatic opinion extraction is a useful technique to monitor consumers'' feedback regarding any particular productsin terms of positive or negative.The management of customer relations can use these feedbacks to improve the products and services and ultimately can make the clients happy. Support Vector Machine (SVM) is one of the most famous and useful supervised machine learning technique used for sentiment classification and opinion extraction. Many extensions and modifications of this algorithm are available today. The purpose of this research is to improve the SVM accuracy through grid search technique for sentiment classificationof textual data.