Live Sentiment Analysis Using Multiple Machine Learning and Text Processing Algorithms
Andrew R Motz, Elizabeth Ranta, Adan Sierra Calderon, Quin Adam, Fadi Alzhouri, Dariush Ebrahimi
Abstract
Due to the massive amount of data being generated on the platform, Twitter has been the subject of numerous sentiment analysis studies. Such social network services generate massive unstructured data streams which make working with them very challenging. The aim of this study is to reliably analyze the sentiment of trending tweets in the Twitter API data stream using a combination of different algorithms to achieve a consensus. The methods we implemented include Support-Vector Machine, Naive Bayes, Textblob, and Lexicon Approach. The hypothesis is that using these methods together would enable us to get more accurate results. Using a labeled dataset to test our model, the results show that the combination of these four algorithms all together performed best with an overall accuracy of 68.29%. We conclude that our combination method of analysis is suitable and fast enough for our data stream and also accurate for analyzing sentiment.