NATURAL LANGUAGE PROCESSING FOR SENTIMENT ANALYSIS IN SOCIAL MEDIA: A DEEP LEARNING APPROACH

Abstract

This study examines the performance of three machine learning models - Convolutional Neural Networks (CNN), Support Vector Machines (SVM), and K-Nearest Neighbors (KNN) - on a dataset of 162,975 tweets. These tweets are categorized as negative, neutral, or positive for Twitter sentiment analysis. The main objective is to assess the accuracy, precision, recall, and F1-score of these models. The results indicate that the CNN model is effective at predicting positive sentiments but struggles with negative and neutral categories, often misclassifying them. Both the SVM and KNN models show similar patterns, with SVM displaying a noticeable bias towards positive predictions and KNN showing a more balanced yet still imperfect performance. The analysis of the confusion matrix underscores the challenge of addressing class imbalances in sentiment analysis tasks. This study emphasizes the importance of developing strategies to improve classification accuracy for underrepresented sentiment classes. By doing so, it contributes to the broader field of Natural Language Processing (NLP) and its applications in automated sentiment analysis. The findings align with existing literature, reinforcing the need for advanced techniques to manage class imbalances in sentiment data.