Researchers and businesses alike are becoming increasingly interested in the automatic study of social network public opinion for a given subject. This work discusses several word embedding methods for sentiment analysis first, and then it gives a summary of the most recent pre-trained models for NLP, which is widely used in sentiment analysis. There is a growing demand for practical ways to manage the enormous volume of multimodal data that is transferred across these channels as more and more recordings become accessible online and in other locations. Currently, its focus is on examining various methods used with various datasets. This approach considers the thoughts and emotions of individuals regarding the occurrence of an incident. Sentiment analysis's main objective is to classify user sentiment into three categories: neutral, negative, and positive. Therefore, we are employing a number of methods such as TEXTBLOB, VADER. We are using a dataset called Twitter dataset, which consists of thousands of tweets. In this some Deep learning (DL) models (like Bag-Of-Words Vectorization Model, LSTM and Transformer Based Models) are used. These models make it way easier to determine the tone of an individual by analysing the tweets in the form of textual data. We will evaluate each accuracy and try to determine which model is appropriate for identifying and analysing sentiments of an individual using the provided dataset. After comparing, we find that the VADER model is best out of all, with an accuracy of 0.64, preventing overfitting and taking manageable amounts of time.