Comparison Of Neural Network and Traditional Classifiers for Twitter Sentiment Analysis

Abstract

Sentiment analysis has been a popular topic of study in the field of social media analysis, particularly when it comes to analyzing the emotions expressed in online comments. This is particularly relevant when it comes to IMDb Movie reviews, where users often express their opinions on the films they have watched. By using sentiment analysis techniques, researchers can gain insights into the overall sentiment of a movie and how it is perceived by the public. This information can be useful for movie studios and producers, as it can help them gauge the success of their films and make decisions about future productions. In this analysis, the performance of neural network models is compared with that of traditional classification methods when applied to the task of sentiment classification of tweets. A dataset of tweets collected from IMDb movie reviews is used for training. Three different models are trained on this dataset: a sequential neural network with two dense layers activated by ReLU and SoftMax functions, logistic regression, and random forest. The performance of these models is evaluated using a variety of metrics, including confusion matrices, AUC graphs, and accuracy and loss curves. It is found that the neural network model achieves an accuracy of approximately 90%, outperforming the logistic regression and random forest models, which achieve accuracies of approximately 90% and 83%, respectively.