In this study, we propose new techniques for feature selection and sentiment analysis using classification algorithms. For this purpose, we collected tweets from 5000 users for a period of one month. We considered the sentiments such as happy, joy, sadness, anger, fear, surprise, distress and disgust and identified the words used to express these features. Based on synonym analysis, we select features for positive sentiments and negative sentiments by proposing a new feature selection algorithm using keyword frequency and semantic analysis. Moreover, we propose a new classification algorithm based on a new type of support vector machine called Group Support Vector Machines (GSVM) to perform major and sub classification of sentiments and to form groups based on the sentiments of people with respect to change in time and location. Finally, the groups are used to form discussion forums on various topics including business, tour, e-learning, religion and sports. The main advantage of the proposed research is to identify people with similar interest based on the sentiments identified from tweets and to form interest groups for discussion on interesting topics. From the experiments conducted in this research, it is observed that the groups formed by sentiment analysis provided >95% accuracy in identifying members for forming interest groups on twitter and hence, is more accurate than the existing systems.
展开▼
