Comparative Study of Twitter Post Classification Based on User Role and Post Type
Today many businesses have started using Twitter as a tool to broadcast and advertise their products and services to their customers. Typically, users would most likely follow persons or companies whose posts are of their interests. Recommending users to follow business, which matches their interests, is an attractive approach to gain more revenue. One possible solution to business recommendation is by applying classification algorithm to predict users Twitter posts into some predefined business categories. Learning characteristics of communication on Twitter is very important for improving classification models. In this paper, we perform a comparative study on user role and post type on the performance of Twiner post classification. We compare between using post-level and user-level data for constructing the classification model. In addition, four different post types are also evaluated in the study. From the evaluation results, the best classification performance is obtained by using user-level data from the users who are direct followers of the business.
Twitter classification social network analysis user role
Chanattha Thongsuk Choochart Haruechaiyasak Somkid Saelee
Faculty of Information Technology King Mongkuts University of Technology North Bangkok Bangkok, THA Human Language Technology Laboratory (HLT) National Electronics and Computer Technology Center (NECT Faculty of Technical Education King Mongkuts University of Technology North Bangkok Bangkok, THAILA
国际会议
海口
英文
338-341
2011-07-15（万方平台首次上网日期，不代表论文的发表时间）