Sentiment Analysis Using Machine Learning Approaches on Social Media Data

Background: Twitter is a microblogging service that allows users to send updates (tweets) to their friends (followers) [1]. It has grown into a substantial sentiment database. Sentiment analysis is the study of emotions and ideas expressed in text [2]. Opinion mining is another name for sentiment analysis. Sentiment analysis of this massively generated data is extremely beneficial for expressing the public's opinion or ideas on a certain issue or trend. Customers can use sentiment analysis to search for products, companies can keep track of and analyse public sentiment about their brands, and it can be used for a variety of other purposes [3].

Objective: To find and classify a person's sentiment towards a certain source of document or text.

Methodology: In this study, we are creating dataset using Twitter API and saving it for experiments. These experiments can be done on variety of public tweet sentiment datasets using ensembles classifier formed various classifiers using majority voting such as Multinomial Naïve Bayesian, Support Vector Machine, Bernoulli Naive Bayes and Logistic Regression can improve classification accuracy and comparison of different algorithms with each other when used alone for sentiment analysis [4].

Result and Discussion: Five machine learning classifiers Logistic Regression, Linear SVC, Multinomial Naive Bayes, Bernoulli Naive Bayes and Ensemble Classifier have been used to obtain the sentiment of twitter data, out of which the results of the Logistic Regression Classifier has obtained 82.92% in all parameters.

Future Work: In the future, we will collect real-time data and use deep learning like CNN, LSTM and their ensemble methods to improve outcomes.