Density-Based Clustering for Twitter Sentiment Analysis Using Artificial Intelligence

In this paper, we tackle the challenge of analyzing Twitter data, which is rich in opinions but lacks labels, making it difficult for computers to organize. We propose a solution that consists of grouping similar tweets together, improving the accuracy of sentiment analysis and helping us better understand people’s opinions and emotions expressed on Twitter. We introduce a new density-based clustering algorithm that identifies dense areas in the dataset by determining the density of each feature in the Term Frequency-Inverse Document Frequency (TF-IDF) data matrix. Next, we use an extended version of the k-means clustering algorithm to generate clustering results based on the identified dense areas. These clustering results categorize tweets into positive, negative, and neutral sentiment within each cluster. We then apply a topic modeling technique, specifically Latent Dirichlet Allocation, for sentiment analysis within each category. Experimental results on synthetic data show that our proposed algorithm outperforms current state-of-the-art approaches. Finally, we present results related to public sentiment regarding the use of ChatGPT, a generative AI model, in education, utilizing Twitter data.