Twitter Sentiment Analysis: A Case of Study from Yachay Tech University

The use of social networks is a trend that has been increasing in recent years. Citizens can express their opinions through social networks, and one of the favorites for this is Twitter due to the nature of the platform. In this social network, people interact with diverse communities through written opinions. However, for this exact reason, the data coming from Twitter is not very structured, which makes it challenging to analyze. Therefore, this paper proposes implementing a model to perform sentiment analysis on Twitter data. For this purpose, the data is first subjected to an extensive pre-processing stage. This step aims to clean the data by eliminating unnecessary information such as emojis, stopwords, etc. Subsequently, the cleaned data is subjected to four tests of different machine learning algorithms. After evaluating the corresponding metrics, it was determined that the best-performing algorithm for this data was Naive Bayes. This algorithm obtained a 69% accuracy rate. Metrics such as precision, recall, f1 score, and a 5-fold cross-validation test were also analyzed.