Machine Learning for Sentiment Analysis Utilizing Social Media

Abstract

Sentimental analysis is a crucial step in natural language processing that aids in figuring out whether a text has a positive, negative, or neutral sentiment. In this experiment, we analyzed the sentiments expressed in tweets that included text, emojis, and emoticons. To categorize the tweets into different sentiments, we utilized four different algorithms: Multinomial Naive Bayes (MNB),Random Forest, Support Vector Machine (SVM) and Decision Tree. In order to increase the model's accuracy, we also combined the predictions from the four algorithms using the Voting Classifier, an ensemble learning technique. To preprocess the data, we used various techniques, such as removing stop words, stemming, and converting emojis and emoticons to their corresponding text representations. The performance of each algorithm was then trained on the preprocessed data using various assessment measures, including accuracy, precision, F1-score and recall. The SVM method fared better than the other algorithms, obtaining an accuracy of 96.27%, according to the data. Furthermore, we applied ensemble learning techniques, such as bagging to improve the performance of all the four algorithms. We also used the Voting Classifier to combine the predictions of the bagging models to further improve the accuracy of the model. The results revealed that the accuracy was increased to 97.21% by combining the bagging and voting classifiers. Overall, the project demonstrates the effectiveness of various algorithms and ensemble learning methods in performing sentimental analysis on tweets containing text, emojis, and emoticons.