Sentiment Analysis on Twitter using Ordinal Regression

With around 186 million daily active users, Twitter is a popular social media platform with increasing user growth every year. The top three twitter markets are in the US, Japan, and India. Users tweet on various topics ranging from sports, politics, finance to entertainment. Twitter is also used by businesses to advertise their products and services. These tweets contain a vast amount of information that can be used in many different ways, from targeting advertisements, understanding customer sentiment, and analyzing users’ opinions about specific people or events. This research aims to apply sentiment analysis on Twitter data by Ordinal Regression. We use the twitter_samples from NLTK Corpus containing positive and negative tweets for our study. We use Random Forest (RF), Support Vector Machine (SVM), and Multinomial Logistic Regression (MLR) classifiers to classify the tweet sentiment into five categories. Results show that this approach outperformed the related work in terms of accuracy. This improvement was possible by applying TextBlob to find the polarity of the tweets and using lemmatization instead of stemming.