Sentiment analysis is a vital part of natural language processing (NLP), and because of its applications in various fields, it has become one of the most famous machine learning (ML) techniques. In this research paper, we have done a comprehensive sentiment analysis of YouTube comments through various ML models like Multi-layer Perceptron (MLP), Naive Bayes, Gradient Boosting, K-Nearest Neighbors (KNN) and Support Vector Machine (SVM). The use of ML models like MLP and Gradient Boosting makes this paper unique because it needs to be explored more in YouTube comments sentiment analysis specifically. Another quality of this paper is the highly achieved accuracy of various models, which is rare in complex problems like sentiment analysis. MLP model achieved the highest accuracy, which is 97.52%. The SVM and Random Forest models also gave an accuracy of above 95%. Gradient Boosting also performed well, with an accuracy of 80.25%. However, the accuracy given by Naive Bayes and KNN was quite low. We used the TextBlob python library to differentiate and categorize the comment text into neutral, positive, or negative comments. Data exploration using techniques like Word Cloud makes it easier to understand the distribution of these texts into different sentiments. The research includes data cleaning and preprocessing, data preparation using Term Frequency and Inverse Document Frequency (TF-IDF), etc. and training of models. It also studies the accuracy and F1-scores achieved in each model and compares them.