To address this, a dataset of YouTube comments labeled for sentiment was acquired. The data was preprocessed by removing stopwords and punctuation, transforming the text into numbers using the TF-IDF method, and various models such as Logistic Regression, Naive Bayes, Decision Tree, and Support Vector Machine (SVM) were trained. The performance of each model was compared, on metrics such as accuracy, precision, recall, and F1-score.
Overall, the project helped demonstrate that with proper precautions in place, even some of the older machine learning algorithms can be effective at determining sentiment in random real-world text like YouTube comments. It also allowed for a better understanding of how important preprocessing really is when working with text data.
]]> sentiment analysis topic became one of the important and essential fields for most. Collection of
people's feelings and sentiment and classifying the data attracted most businesses and companies.
Recently, twitter sentiment analysis has attracted much attention, because of Twitter's growth and
popularity. The solution for handling enormous amounts of data from social media is a new term
called Big data. Big data is not just for having a large amount of data, but also the importance of
processing and the usage of the data.]]>
To address these questions, throughout the research, traditional machine learning models including Naive Bayes, Linear Support Vector Classification (SVC) and Logistic Regression were implemented. Among these approaches, Linear SVC achieved the best scores across all evaluation metrics. In the deep learning category, Long Short-Term Memory (LSTM) networks were applied. This approach provided exceptional performance which was overall better than traditional models. RNNs scored similarly as Linear SVC, while outperforming other traditional algorithms.

A hybrid approach including the BERT model was another method that was explored, which combined specific architecture with deep learning-based contextual understanding. The results demonstrated high classification results, which supports the hypothesis that hybrid models can increase performance of sentiment prediction. Furthermore, fine-tuning of different models improved their performance, which highlights the importance of optimizing pretrained models for specific types of analysis.

Overall, the findings confirm that deep learning models usually outperform traditional variants of machine learning methods while hybrid models can offer additional potential and perspective for enhancing sentiment classification in news articles. The study provides deep and valuable insights into effectiveness of different sentiment analysis and natural language processing (NLP) techniques, while at the same time discussing new possibilities and improvements in the field.]]>