Performance of Machine Learning Models in Predicting Sentiments of Post-Covid Patients

Abstract

With the widespread use of social media platforms, sentiment analysis of user-generated content has become a crucial task in understanding public opinion and trends. In this paper, we compare the performance of three popular machine learning models, namely Random Forest, Support Vector Machine (SVM), and Logistic Regression, in predicting sentiments of post-COVID patients on social media tweets. The study utilizes a dataset of labeled tweets representing positive, negative, and neutral sentiments. The preprocessing of textual data involves tokenization, stop-word removal, and conversion to lowercase to create a suitable input for the models. We utilize Term Frequency-Inverse Document Frequency (TF-IDF) vectorization to transform the text data into numerical features. The sentiment labels are converted to numeric representations for model training and evaluation. The three machine learning models are trained and evaluated on the dataset using metrics such as accuracy, precision, recall and F1-score. The evaluation results are presented and analyzed for each model, providing insights into their strengths and weaknesses in predicting sentiments. The experimental results demonstrate that Random Forest achieves the highest accuracy and F1-score, closely followed by SVM, while Logistic Regression performs slightly lower in comparison. However, all three models exhibit strong predictive capabilities, and their performances vary depending on the specific sentiment class. The findings provide valuable information for researchers and practitioners seeking to employ sentiment analysis in social media monitoring and other related applications. Overall, this study contributes to the understanding of the capabilities of Random Forest, SVM, and Logistic Regression models in sentiment analysis of social media tweets, and offers valuable insights for selecting the most suitable model for specific sentiment prediction tasks.