Analyzing user behaviors based on temporal patterns of sequential pattern evaluation indices on twitter

Abstract

With social media sites, such as Twitter, providing a visual record of the daily interests and concerns of users in the form of tweets and tweeting behaviors, there is growing demand among users, such as corporations, to identify other interested users. However, accurately determining whether users who receive information (such as tweets) from enterprise users have a genuine interest in it can be difficult. In this study, the user behavior of resending information received on Twitter (retweeting) is analyzed with the aim of developing a method for constructing a model for predicting retweeting behavior using the content of past tweeting history via evaluation indices of words and phrases in the users’ tweets. This paper analyzes the tweets sent by large online retail websites and by the followers who receive them, comparing the feature words obtained from the retweets with those in the tweets sent by the followers. This paper also discusses the feasibility of constructing a behavior prediction model by extracting temporal patterns of evaluation indices that are created from the usage frequencies of feature words and phrases obtained from followers’ tweets.