Sentiment analysis of twitter texts using machine learning algorithms

In this thesis, sentiment analysis as the use of natural language processing and machine learning classifiers have been studied on Trump's tweets scraped web page, which is saved in the form of dataset. After data preparation, the most important sentiment analysis procedures have been applied to the host dataset. Also, other natural language processing strategies have been processed, like cleaning the dataset in order to be ready for text vectorization. In cleaning the textual data, all the required techniques like removing stopwords, word lemmatization, regular expression, and tokenization have been used to remove undesired words. We succeeded in reducing the size of the "content" feature in the dataset with the target of taking fewer capacity. Since the two last decades with the development of social media networks, hateful activities have become a phenomenon, this became a challenging task to know the subjective polarities of each one's published text; therefore, each sentence has been judged-on regarding their polarities whether they are positive, negative or neutral. At the end, by using machine learning algorithms like (Random Forest classifier, Gaussian Naive Bayes, and Support Vector Machine), the cleaned data has been trained and tested to see the accuracy of the prediction results, the comparison shows 88%, 72%, and 89% respectively for each classifier.