Development of machine learning based methods for social sentiment classification from brief texts

With the development of technology, human life enters the virtual world more and more. With the development of the Internet, social media networks such as Twitter, Facebook, Instagram, Tumblr, Google+, etc. have become indispensable parts of human life. In these networks, where millions of messages are circulating in one day, people share their lives with their friends, family and even people they do not know and are happy with it. The sharing of what people think about specific topics and products and what they like and what they need is starting to become interesting for the trade, manufacturing and service sectors. This has made important Text Mining and Text Analysis as well as the Sentiment Analysis. In the first study, classification of news texts with different feature extraction methods and term weighting methods which based machine learning methods has been important in terms of testing the efficiency and success of the methods. With the second study, two different data sets including Twitter posts are also classified by positive, negative, and neutral classes and the nearest neighbor algorithm based on particle swarm optimization, by conducting sentiment analysis with machine learning techniques. It has been observed that the proposed method yields more successful results when compared to the cuckoo algorithm proposed previously on the same data sets.